大语言模型上下文工程综述
Lingrui Mei1,6,† Jiayu Yao1,6,† Yuyao Ge1,6,† Yiwei Wang2 Baolong Bi1,6,†
Yujun Cai3 Jiazhi Liu1 Mingyu Li1 Zhong-Zhi Li6 Duzhen Zhang6
Chenlin Zhou4 Jiayi Mao5 Tianze Xia6 Jiafeng Guo1,6,† Shenghua Liu1,6,†,
1 Institute of Computing Technology, Chinese Academy of Sciences（中国科学院计算技术研究所）,
2 University of California, Merced, 3 The University of Queensland,
4 Peking University（北京大学）, 5 Tsinghua University（清华大学）,
6 University of Chinese Academy of Sciences（中国科学院大学）
arXiv:2507.13334v2 [cs.CL] 21 Jul 2025
摘要： 大语言模型（Large Language Models, LLMs）的性能从根本上取决于推理（inference）时所提供的上下文信息。本综述引入上下文工程（Context Engineering）这一形式化学科——它超越简单的提示设计，涵盖面向 LLMs 的信息载荷（information payloads）的系统性优化。我们提出一套全面的分类体系（taxonomy），将上下文工程分解为基础组件（Foundational Components）以及将其整合为智能系统的复杂实现（Implementations）。我们首先考察基础组件：（1）上下文检索与生成（Context Retrieval and Generation），涵盖基于提示的生成与外部知识获取；（2）上下文处理（Context Processing），面向长序列处理、自我精炼（self-refinement）与结构化信息整合；（3）上下文管理（Context Management），涵盖记忆层次（memory hierarchies）、压缩（compression）与优化。随后，我们探讨这些组件如何在架构上被整合，以构建复杂的系统实现（System Implementations）：（1）检索增强生成（Retrieval-Augmented Generation, RAG），包括模块化（modular）、智能体式（agentic）与图增强（graph-enhanced）架构；（2）记忆系统（Memory Systems），支持持久化交互；（3）工具集成推理（Tool-Integrated Reasoning），用于函数调用（function calling）与环境交互；（4）多智能体系统（Multi-Agent Systems），协调通信与编排（orchestration）。通过对逾 1400 篇研究论文的系统分析，本综述不仅为该领域确立了技术路线图，也揭示了一项关键研究缺口：模型能力之间存在根本性的不对称。尽管当前模型在先进上下文工程的加持下，在理解复杂上下文方面展现出卓越能力，但在生成同等精巧、长篇幅输出方面却表现出明显局限。弥补这一差距是未来研究的核心优先事项。最终，本综述为推进上下文感知人工智能的研究者与工程师提供了统一框架。
† 同时隶属：(1) Key Laboratory of Network Data Science and Technology, ICT, CAS；(2) State Key Laboratory of AI Safety
通讯作者（Corresponding Author）
关键词： 上下文工程，大语言模型，LLM 智能体，多智能体系统
日期：July 21, 2025
代码仓库：https://github.com/Meirtz/Awesome-Context-Engineering
联系方式：meilingrui25b@ict.ac.cn, liushenghua@ict.ac.cn
目录
- 1 引言
- 2 相关工作
- 3 为何需要上下文工程？
- 3.1 上下文工程的定义
- 3.2 为何需要上下文工程
- 3.2.1 当前局限
- 3.2.2 性能提升
- 3.2.3 资源优化
- 3.2.4 未来潜力
- 4 基础组件
- 4.1 上下文检索与生成
- 4.1.1 提示工程与上下文生成
- 4.1.2 外部知识检索
- 4.1.3 动态上下文组装
- 4.2 上下文处理
- 4.2.1 长上下文处理
- 4.2.2 上下文自我精炼与适应
- 4.2.3 多模态上下文
- 4.2.4 关系型与结构化上下文
- 4.3 上下文管理
- 4.3.1 基本约束
- 4.3.2 记忆层次与存储架构
- 4.3.3 上下文压缩
- 4.3.4 应用
- 4.1 上下文检索与生成
- 5 系统实现
- 5.1 检索增强生成
- 5.1.1 模块化 RAG 架构
- 5.1.2 智能体式 RAG 系统
- 5.1.3 图增强 RAG
- 5.1.4 应用
- 5.2 记忆系统
- 5.2.1 记忆架构
- 5.2.2 记忆增强智能体
- 5.2.3 评估与挑战
- 5.3 工具集成推理
- 5.3.1 函数调用机制
- 5.3.2 工具集成推理
- 5.3.3 智能体–环境交互
- 5.4 多智能体系统
- 5.4.1 通信协议
- 5.4.2 编排机制
- 5.4.3 协调策略
- 5.1 检索增强生成
- 6 评估
- 6.1 评估框架与方法
- 6.1.1 组件级评估
- 6.1.2 系统级集成评估
- 6.2 基准数据集与评估范式
- 6.2.1 基础组件基准测试
- 6.2.2 系统实现基准测试
- 6.3 评估挑战与新兴范式
- 6.3.1 方法局限与偏差
- 6.3.2 新兴评估范式
- 6.3.3 安全与鲁棒性评估
- 6.1 评估框架与方法
- 7 未来方向与开放挑战
- 7.1 基础研究挑战
- 7.1.1 理论基础与统一框架
- 7.1.2 扩展定律与计算效率
- 7.1.3 多模态整合与表示
- 7.2 技术创新机遇
- 7.2.1 下一代架构
- 7.2.2 高级推理与规划
- 7.2.3 复杂上下文组织与图问题求解
- 7.2.4 智能上下文组装与优化
- 7.3 应用驱动的研究方向
- 7.3.1 领域专门化与适应
- 7.3.2 大规模多智能体协调
- 7.3.3 人机协作与集成
- 7.4 部署与社会影响考量
- 7.4.1 可扩展性与生产部署
- 7.4.2 安全、安保与鲁棒性
- 7.4.3 伦理考量与负责任开发
- 7.1 基础研究挑战
- 8 结论
1. 引言
大语言模型（LLMs）的出现标志着人工智能的范式转变，在自然语言理解、生成与推理方面展现出前所未有的能力 [103, 1067, 459]。然而，这些模型的性能与效力从根本上受其所接收上下文的制约。这一上下文——从简单的指令性提示（prompt）到复杂的外部知识库——是引导其行为、增强其知识并释放其能力的主要机制。随着 LLMs 从基础的指令遵循系统演进为复杂应用的核心推理引擎，设计与管理其信息载荷的方法也相应演进为上下文工程这一形式化学科 [25, 1265, 1068]。
上下文工程的研究版图以爆炸性速度扩张，催生了大量专门化却彼此割裂的研究领域。我们将这一版图概念化为由基础组件及其后续实现所构成。基础组件通过三个关键阶段表征上下文工程的系统化流水线：上下文检索与生成，涵盖基于提示的生成与外部知识获取 [25, 597, 48]；上下文处理，涉及长序列处理、自我精炼机制与结构化信息整合 [200, 741, 495]；以及上下文管理，面向记忆层次、压缩技术与优化策略 [1372, 1082, 819]。
这些基础组件构成了更复杂、面向应用的实现的基石，从而将 LLMs 与外部现实相桥接。这类系统包括：已演进为模块化与智能体式架构、实现动态知识注入的先进检索增强生成（RAG） [597, 316, 973, 315]；模拟人类认知官能以实现信息持久保留的显式记忆系统 [1191, 943, 1372]；以及整个智能智能体系统（Intelligent Agent Systems）生态。后一类代表了上下文工程的巅峰——智能体（agent）借助函数调用与工具集成推理与世界交互 [939, 864, 669]，并依赖精巧的智能体通信协议与上下文编排（Context Orchestration），在多智能体配置中达成复杂目标 [360, 250, 902, 128]。
尽管上述各领域均产生了大量创新，但它们大多被孤立地研究。这种碎片化发展遮蔽了各技术之间的根本联系，既给试图理解更广版图的研究者、也给希望有效运用这些方法的实践者带来显著障碍。该领域迫切需要一个统一框架，系统组织这些多样化技术，澄清其底层原理，并阐明其间的相互依存关系。
为填补这一关键空白，本综述提供首份面向 LLMs 上下文工程的全面、系统评述。我们的主要贡献是一套新颖的结构化分类体系，对设计、管理与优化上下文的多面技术加以分类。该分类体系将领域组织为连贯类别，区分基础组件及其向复杂系统实现的整合。通过这一框架，我们：（1）对各领域的前沿现状给出清晰、结构化的概览；（2）分析不同方法的核心机制、优势与局限；（3）识别总体挑战并勾勒有前景的未来研究方向。本工作既可作为导航上下文工程复杂版图的技术路线图，也可作为深化理解与催化未来创新的基础。
本文其余部分组织如下。在讨论相关工作并形式化定义上下文工程之后，我们首先考察该领域的基础组件，涵盖上下文检索与生成、上下文处理与上下文管理。随后探讨其系统实现，包括检索增强生成、记忆系统、工具集成推理与多智能体系统。最后讨论评估方法、未来研究方向并给出结论。图 1 给出了我们分类体系的全面概览，展示了上下文工程版图中各技术的层次组织及其相互关系。
2. 相关工作
LLMs 的迅速成熟催生了大量旨在勾勒其多面版图的综述文献。既有工作虽有价值，但大多聚焦于我们所谓上下文工程这一更广领域中的特定纵向子域。本综述力图通过提供横向、统一的分类体系来补充这些努力——该体系区分基础组件及其向复杂系统的整合，从而桥接这些专门领域。
基础组件 众多综述已讨论构成有效上下文操控核心技术能力的上下文工程基础组件。上下文检索与生成的挑战同时涵盖提示工程（prompt engineering）方法与外部知识获取技术。关于提示工程的综述已系统梳理引导 LLM 行为的大量技术，从基础少样本（few-shot）方法到高级结构化推理框架 [25, 257, 1322]。外部知识检索与整合技术——尤其是通过知识图谱与结构化数据源——则在综述表示技术、整合范式及其在增强 LLMs 事实接地（grounding）方面应用的工作中得到评述 [489, 432, 823, 897]。
图 1：大语言模型中上下文工程的分类体系，划分为基础组件、系统实现、评估方法与未来方向。各领域涵盖具体技术与框架，共同推进面向 LLMs 的信息载荷系统性优化。（原文图见 PDF 第 6 页）
上下文处理领域则面向处理长序列、自我精炼机制与结构化信息整合的技术挑战。长上下文（Long Context）处理在分析扩展上下文窗口、优化注意力（attention）机制与高效管理记忆等技术的综述中得到讨论 [837, 651, 1298, 272]。LLMs 的内部认知过程也日益成为综述对象，关于自我情境化技术与自我改进范式的工作日益突出 [1339, 231, 1176, 943]。
最后，上下文管理文献聚焦于记忆层次、压缩技术与优化策略，使信息能够在计算约束下得到有效组织与检索。尽管专门将上下文管理作为统一领域加以全面综述的工作仍属有限，但有关记忆系统与上下文压缩技术的相关工作为这些关键能力提供了基础性洞见。
系统实现 与此并行，文献已广泛覆盖将基础组件整合为复杂架构、以满足现实应用需求的系统实现。RAG 领域受到大量关注，基础性综述追溯了其发展及其对缓解幻觉的影响 [315, 257, 1140]。近期工作则综述了向模块化、智能体式与图增强 RAG 架构的演进 [166, 628, 120, 316, 1401]。
支持持久交互与认知架构的记忆系统，已通过聚焦记忆增强智能体及其应用的综述加以探讨。更广的基于 LLM 的智能体类别构成另一基础领域，已有工作对自主智能体、其架构、规划与方法进行了全面概述 [1099, 725, 281, 849, 1350, 504, 1281]。
涵盖函数调用机制与智能体–环境交互的工具集成推理也有充分文献记载，探讨了从单工具系统到复杂编排框架的演进 [669, 864, 777, 875]。向多智能体系统（MAS）的演进是另一焦点，相关综述详述了 MAS 工作流、基础设施、通信协议与协调机制 [631, 360, 250, 1244, 38, 509, 191, 464]。
评估 评估这些复杂系统的关键方面已得到充分评述，相关工作分析了用于评估组件级与系统级能力及性能的基准测试（benchmark）与方法 [1268, 384, 841, 314]。该评估文献横跨基础组件评估与集成系统评估两类范式。
我们的贡献 尽管上述综述对其各自领域提供了不可或缺的深入分析，但它们本质上呈现的是该领域的碎片化图景。RAG 作为外部记忆的一种形式、工具使用作为上下文获取方法、以及提示工程作为编排这些组件的语言——其间的联系往往被隐而不宣。本工作的独特之处在于提出上下文工程作为统一抽象，显式区分基础组件与其在复杂实现中的整合。通过将这些分散领域组织到单一、连贯的分类体系中，本综述旨在阐明其间的根本关系，为上下文如何被生成、处理、管理并用于引导下一代智能系统提供整体地图。
3. 为何需要上下文工程？
随着大语言模型（LLMs）从简单的指令遵循系统演进为复杂、多面应用的核心推理引擎，与之交互的方法也必须随之演进。“提示工程”一词虽具基础性，但已不足以涵盖设计、管理与优化现代 AI 系统所需信息载荷的全部范围。这些系统并不运行于单一、静态的文本字符串之上；它们依赖动态、结构化且多面的信息流。为此，我们引入并形式化上下文工程这一学科。
图 2：上下文工程演进时间线：全面可视化 2020 至 2025 年间上下文工程实现的发展轨迹，展示从基础 RAG 系统到复杂多智能体架构与工具集成推理系统的演进。（原文图见 PDF 第 8 页）
3.1. 上下文工程的定义
为形式化定义上下文工程，我们从自回归 LLM 的标准概率模型出发。模型由参数 θ 参数化，在给定输入上下文 C 时，通过最大化条件概率生成输出序列 Y = (y₁, …, y_T)：
$$P_\theta(Y \mid C) = \prod_{t=1}^{T} P_\theta(y_t \mid y_{<t}, C) \tag{1}$$
历史上，在提示工程范式中，上下文 C 被当作整体、静态的文本字符串，即 C = prompt。这一观点对现代系统而言已不充分。
上下文工程将上下文 C 重新概念化为由信息组件 c₁, c₂, …, cₙ 构成的动态结构化集合。这些组件由一组函数进行来源获取、过滤与格式化，最终由高层组装函数 A 加以编排：
$$C = A(c_1, c_2, \ldots, c_n) \tag{2}$$
组件 cᵢ 并非任意；它们直接映射到本综述的核心技术领域：
- $c_{\text{instr}}$：系统指令与规则（上下文检索与生成，第 4.1 节）。
- $c_{\text{know}}$：外部知识，通过 RAG 等函数检索或来自集成知识图谱（RAG，第 5.1 节；上下文处理，第 4.2 节）。
- $c_{\text{tools}}$：可用外部工具的定义与签名（函数调用与工具集成推理，第 5.3 节）。
- $c_{\text{mem}}$：来自先前交互的持久信息（记忆系统，第 5.2 节；上下文管理，第 4.3 节）。
- $c_{\text{state}}$：用户、世界或多智能体系统的动态状态（多智能体系统与编排，第 5.4 节）。
- $c_{\text{query}}$：用户的即时请求。
上下文工程的优化问题。 由此视角看，上下文工程即是寻找理想的上下文生成函数集合（我们将其统记为 $F = {A, \text{Retrieve}, \text{Select}, \ldots}$）以最大化 LLM 输出期望质量的形式化优化问题。给定任务分布 $T$，目标为：
$$F^{} = \arg\max_{F} \mathbb{E}{\tau \sim T}\bigl[\text{Reward}\bigl(P{\theta}(Y \mid C_{F}(\tau)), Y_{\tau}^{}\bigr)\bigr] \tag{3}$$
其中 $\tau$ 为特定任务实例，$C_{F}(\tau)$ 为由 $F$ 中函数为该任务生成的上下文，$Y_{\tau}^{*}$ 为真值或理想输出。该优化受硬约束制约，最突出的是模型的上下文长度限制 $|C| \leq L_{\max}$。
数学原理与理论框架。 这一形式化揭示了更深层次的数学原理。组装函数 $A$ 是一种动态上下文编排（Dynamic Context Orchestration）形式，即由格式化与拼接操作构成的流水线 $A = \text{Concat} \circ (\text{Format}{1}, \ldots, \text{Format}{n})$，其中每个函数都必须针对 LLM 的架构偏置（例如注意力模式）进行优化。
知识检索 $c_{\text{know}} = \text{Retrieve}(\ldots)$ 可被框定为信息论最优性问题。目标是在给定查询 $c_{\text{query}}$ 的条件下，选择与目标答案 $Y^{*}$ 互信息最大的知识：
$$\text{Retrieve}^{} = \arg\max_{\text{Retrieve}} I(Y^{}; c_{\text{know}} \mid c_{\text{query}}) \tag{4}$$
这确保检索到的上下文不仅语义相似，而且对求解任务具有最大信息量。
进一步地，整个过程可通过贝叶斯上下文推理（Bayesian Context Inference）的视角理解。我们并非确定性地构造上下文，而是推断最优上下文后验 $P(C \mid c_{\text{query}}, \text{History}, \text{World})$。由贝叶斯定理，该后验正比于给定上下文下查询的似然以及上下文相关性的先验概率：
$$P(C \mid c_{\text{query}}, \ldots) \propto P(c_{\text{query}} \mid C) \cdot P(C \mid \text{History}, \text{World}) \tag{5}$$
决策论目标则是寻找在可能答案分布上最大化期望奖励的上下文 $C^{*}$：
$$C^{} = \arg\max_{C} \int P(Y \mid C, c_{\text{query}}) \cdot \text{Reward}(Y, Y^{}), dY \cdot P(C \mid c_{\text{query}}, \ldots) \tag{6}$$
这一贝叶斯表述为处理不确定性、通过更新先验进行自适应检索，以及在多步推理任务中维持关于上下文的信念状态提供了原则性方法。
| 维度 | 提示工程 | 上下文工程 |
|---|---|---|
| 模型 | $C = \text{prompt}$（静态字符串） | $C = A(c_{1}, c_{2}, \ldots, c_{n})$（动态、结构化组装） |
| 目标 | $\arg\max_{\text{prompt}} P_{\theta}(Y \mid \text{prompt})$ | $F^{} = \arg\max_{F} \mathbb{E}{\tau \sim T}[\text{Reward}(P{\theta}(Y \mid C_{F}(\tau)), Y_{\tau}^{})]$ |
| 复杂性 | 在字符串空间上手动或自动搜索 | 对 $F = {A, \text{Retrieve}, \text{Select}, \ldots}$ 的系统级优化 |
| 信息 | 信息内容固定于提示词之内 | 在约束 $|C| \leq L_{\max}$ 下最大化任务相关信息 |
| 状态 | 主要为无状态 | 内在有状态，具有 $c_{\text{mem}}$ 与 $c_{\text{state}}$ 的显式组件 |
| 可扩展性 | 脆弱性随长度与复杂性增加 | 通过模块化组合管理复杂性 |
| 误差分析 | 人工检查与迭代精炼 | 对各个上下文函数的系统评估与调试 |
表 1： 提示工程与上下文工程范式的比较。
范式比较。 上下文工程的形式化凸显了其与传统提示工程的根本区别。上表总结了关键差异。
总之，上下文工程提供了构建、理解与优化那些正日益定义该领域未来的复杂、上下文感知 AI 系统所需的形式化、系统化框架。它将焦点从提示设计的“艺术”转向信息物流与系统优化的“科学”。
上下文扩展（Context Scaling）。 上下文扩展涵盖两个共同界定上下文信息处理范围与精密度的基本维度。第一维是长度扩展（length scaling），应对处理超长序列的计算与架构挑战，将上下文窗口从数千扩展至数百万 token，同时在扩展的叙事、文档与交互中保持连贯理解。这涉及复杂的注意力机制、记忆管理技术以及架构创新，使模型能够在大幅扩展的输入序列上维持上下文连贯性。
第二维同样关键，即多模态与结构扩展（multi-modal and structural scaling），将上下文从简单文本扩展到涵盖多维、动态、跨模态的信息结构。这包括时间上下文（理解时间依赖关系与序列）、空间上下文（解释基于位置与几何的关系）、参与者状态（跟踪多个实体及其演化条件）、意图上下文（理解目标、动机与隐含目的）以及文化上下文（在特定社会与文化框架内解释交流）。
现代上下文工程必须同时应对这两个维度，因为现实世界应用要求模型不仅处理冗长文本信息，还要处理多样数据类型，包括结构化知识图谱、多模态输入（文本、图像、音频、视频）、时间序列以及人类自然理解的隐含上下文线索。这种上下文扩展的多维方法代表了从参数扩展向开发能够理解复杂、模糊上下文之系统的根本转变，这些系统反映了人类智能在面对复杂世界时的细微本质 [1044]。
3.2. 为什么需要上下文工程
3.2.1. 当前局限
大语言模型面临亟待复杂上下文工程方法的关键技术障碍。自注意力机制随序列长度增加施加二次计算与内存开销，对处理扩展上下文构成重大障碍，并显著影响聊天机器人与代码理解模型等现实应用 [1025, 985]。商业部署通过重复的上下文处理进一步加剧这些挑战，引入额外延迟与基于 token 的定价成本 [1025]。
除计算约束外，LLM 还表现出令人担忧的可靠性问题，包括频繁幻觉、对输入上下文的不忠实、对输入变化的问题性敏感，以及表面上句法正确却缺乏语义深度或连贯性的响应 [959, 1288, 529]。
提示工程过程呈现方法论挑战：其近似驱动且主观的方法狭隘地聚焦于任务特定优化，而忽视个体 LLM 行为 [806]。尽管存在这些挑战，通过精确且上下文丰富的提示词减少歧义并增强响应一致性，提示工程对有效利用 LLM 仍然至关重要 [972]。
3.2.2. 性能提升
上下文工程通过检索增强生成(RAG)与叠加提示（superposition prompting）等技术带来实质性性能提升，已记录的改进包括文本导航准确率提高 18 倍、94% 的成功率，以及跨专业领域通过仔细构造提示词与自动优化获得的显著收益 [271, 774, 687]。
结构化提示技术——尤其是思维链(CoT)方法——通过中间步骤实现复杂推理，同时增强元素感知摘要能力，以整合源文档中的细粒度细节 [1147, 756, 1129]。通过精心选择演示示例实现的少样本学习带来实质性性能增益，包括代码摘要 BLEU-4 分数提高 9.90%，以及缺陷修复精确匹配指标提高 175.96% [310]。
领域特定的上下文工程在专业应用中尤为有价值：执行感知调试框架在代码生成基准测试上实现高达 9.8% 的性能提升，硬件设计应用则受益于专门的测试台生成与安全属性验证 [1370, 881, 44]。这些针对性方法弥合了通用模型训练与专业领域需求之间的差距。
3.2.3. 资源优化
上下文工程通过智能内容过滤以及经由精心设计的提示词直接传递知识，为资源密集型传统方法提供了高效替代方案 [636, 676]。即使相关信息从输入上下文中被删除，LLM 仍可生成期望响应，利用上下文线索与先验知识在保持响应质量的同时优化上下文长度使用，这在数据获取挑战显著的领域尤为宝贵 [636, 676]。
专门的优化技术进一步增强效率收益：上下文感知与责任调优显著降低 token 消耗，动态上下文优化采用精确的 token 级内容选择，以及用于长上下文推理的注意力引导机制 [430, 952, 354]。这些方法在最大化信息密度的同时降低处理开销并维持性能质量 [952, 354]。
3.2.4. 未来潜力
上下文工程通过上下文学习(ICL)实现灵活适应机制，使模型无需显式重训练即可适应新任务，上下文窗口大小直接影响可用于任务适应的示例数量 [623]。先进技术整合压缩与选择机制以进行高效模型编辑，同时保持上下文连贯性 [625]。这种适应性在低资源场景中尤为有价值，使其能够在不需要领域特定微调的情况下有效运用各类提示工程技术，包括零样本方法、少样本示例与角色上下文 [932, 129, 1083]。
包括上下文学习(ICL)、思维链(CoT)、思维树（tree-of-thought）与规划方法在内的复杂上下文工程技术，为细腻的语言理解与生成能力奠定基础，同时优化检索与生成过程，以支撑稳健、上下文感知的 AI 应用 [803, 982]。
未来研究方向表明，通过带有 logit 对比机制的思维链增强 [961]、更好地跨领域利用不同类型上下文——尤其是在结合语法、语义、执行流与文档的代码智能任务中 [1102]，以及随着先进语言模型持续证明提示工程的持久价值而理解最优上下文利用策略 [1087]，在推进上下文敏感应用方面具有巨大潜力。向复杂过滤与选择机制演进是应对 Transformer 架构扩展局限同时保持性能质量的关键路径。
4. 基础组件（Foundational Components）
上下文工程建立在三个基础组件之上，它们共同应对大语言模型中信息管理的核心挑战：上下文检索与生成（Context Retrieval and Generation）通过提示工程、外部知识检索与动态上下文组装获取适当的上下文信息；上下文处理（Context Processing）通过长序列处理、自我精炼机制与结构化数据集成对已获取信息进行变换与优化；上下文管理（Context Management）通过应对基本约束、实现复杂记忆层次结构并开发压缩技术，解决上下文信息的高效组织与利用。这些基础组件为所有上下文工程实现确立了理论与实践基础，构成一个综合框架：各组件分别应对上下文工程流水线的不同方面，同时保持协同关系，以实现全面的上下文优化与有效的上下文工程策略。
4.1. 上下文检索与生成
上下文检索与生成构成上下文工程的基础层，涵盖为 LLM 系统检索与构建相关信息。该组件通过三种主要机制应对获取适当上下文信息的关键挑战：基于提示的生成，用于设计有效指令与推理框架；外部知识检索，用于访问动态信息源；以及动态上下文组装，将已获取组件编排为连贯、任务优化的上下文。
图 3： 上下文工程框架：上下文工程组件的综合分类体系，包括上下文检索与生成、上下文处理与上下文管理，并集成到检索增强生成(RAG)系统、记忆架构、工具集成推理与多智能体协调机制等系统实现中。
4.1.1. 提示工程与上下文生成
提示工程与上下文生成构成上下文检索的基础层，涵盖将艺术与科学相结合以为 LLM 设计有效指令的策略性输入设计。CLEAR 框架——简洁性（conciseness）、逻辑性（logic）、明确性（explicitness）、适应性（adaptability）与反思性（reflectiveness）——指导有效的提示词构造，而核心架构则整合任务指令、上下文信息、输入数据与输出指示符 [708, 1142, 575, 213, 25]。
零样本与少样本学习范式。 零样本提示使模型无需先验示例即可执行任务，完全依赖指令清晰度与预训练知识 [1371, 340, 559, 67, 1054]。少样本提示通过纳入有限示例来扩展这一能力以引导模型响应，经由策略性示例选择演示任务执行 [1371, 405, 103, 552, 794, 1381]。上下文学习(ICL)通过在提示词中利用演示示例，使模型无需参数更新即可适应新任务，其性能显著受示例选择与排序策略影响 [369, 103, 1296, 1024, 928, 852, 1148, 352, 582]。
思维链基础。 思维链(CoT)提示将复杂问题分解为中间推理步骤，映照人类认知 [1147, 405, 340, 947, 609]。零样本 CoT 使用如“Let’s think step by step”等触发短语，将 MultiArith 准确率从 17.7% 提升至 78.7% [559, 1107, 478, 668]，Automatic Prompt Engineer 的精炼可带来额外增益 [1224, 532]。
思维树（Tree-of-Thoughts, ToT）将推理组织为具有探索、前瞻与回溯能力的层次结构，将 Game of 24 成功率从 4% 提升至 74% [1255, 221, 563, 604]。思维图（Graph-of-Thoughts, GoT）将推理建模为任意图，以思想为顶点、依赖为边，相较 ToT 质量提升 62%，成本降低 31% [69, 832, 1376]。
认知架构集成。 核心思想：将人类解题时的心智操作（澄清目标、拆分、筛选、抽象等）显式写入上下文，使 LLM 按可编排的认知步骤推理，而非一次性“想到哪写到哪”。相对 CoT 的自由中间步骤，认知架构强调操作有名、可复用、可编排 [564, 563, 1214, 1173]。
（1）认知提示（Cognitive Prompting）。 将推理拆为固定认知操作（COP）：目标澄清 → 分解 → 过滤 → 重组 → 模式识别 → 抽象 → 泛化 → 整合。三种调度变体——确定性（固定顺序）、自适应（模型自选下一步）、混合（正确轨迹作少样本 CoT）——在 GSM8K 上均优于裸问答；LLaMA3.1-70B 准确率约从 0.87（无提示）→ 0.89（静态）→ 0.91（反思式自适应）[564]。
学习示例（算术题，静态 COP 节选）：
问题： John 以 60 mph 开 3 小时后掉头，想用 4 小时到家；前 2 小时堵死不动，接着 0.5 小时以 30 mph，剩余时间以 80 mph。4 小时结束时离家多远？
- 目标澄清： 求 4 小时返程后的剩余距离（非总路程）。
- 分解 / 过滤： 去程距离 (=60\times3=180)；返程分段：(0+30\times0.5+80\times1.5=135)。
- 整合： (180-135=45) → 45 英里。
关键不在“算对”，而在每一步对应一种可命名的认知操作，推理轨迹可检查、可复用。
（2）Guilford 智力结构（SOI）提示。 用心理学三维网格约束“怎么想”：操作（认知 / 记忆 / 发散 / 聚合 / 评价）× 内容（语义 / 符号 / 图形…）× 产物（单元 / 关系 / 系统 / 蕴涵…）。模型先选格再作答，强制结构化而非自由发挥 [562, 195]。
学习示例（政策分析）：
问题： 技术变革与全球化如何冲击传统产业？
选格：操作=认知，内容=语义，产物=关系 → 输出聚焦“技术→创业冲击→传统产业应对→政策权衡”的因果链，避免散乱罗列。
（3）认知工具（Cognitive Tools）。 进一步把操作做成可调用模块（agentic tool-calling）：understand_question、recall_related、examine_answer、backtracking——主模型编排，子调用仍由同一 LLM 执行，但彼此隔离，降低步骤间干扰 [247]。GPT-4.1 在 AIME2024 上 pass@1 从 **26.7% → 43.3%**（接近 o1-preview 的 44.6%），无需额外 RL 训练 [247, 1038]。
学习示例（竞赛数学，工具编排）：
understand_question：抽出已知量、未知量与可用定理；- 尝试求解；卡住则
recall_related：唤起相似题型与解法；examine_answer：自检计算与约束遗漏；- 发现错误则
backtracking：退回错误步骤并换路径。
| 层次 | 形态 | 一句话 |
|---|---|---|
| 认知提示 | 提示词中的操作序列 | 告诉模型“按哪些心智步骤想” |
| SOI 提示 | 操作×内容×产物选格 | 用心理学分类约束推理维度 |
| 认知工具 | 可调用的推理模块 | 把心智步骤做成隔离的工具 API |
要义： 上下文工程在此不只堆信息，而是把认知过程本身装配进上下文——从“一步写完”到“可编排的心智流水线”。
代码示例（教学）：
论文/prompt-techniques-glm已落地cognitive_prompting（含 adaptive/hybrid）、soi_prompting、cmt_prompting、cognitive_tools、thinking_intervention、got。
运行：PYTHONPATH=src python scripts/run.py --technique cognitive_prompting
| 方法 | 描述 |
|---|---|
| Self-Refine [741, 924] | 使 LLM 通过迭代反馈与精炼循环改进输出，同一模型充当生成器、反馈提供者与精炼器，无需监督训练。 |
| Multi-Aspect Feedback [805] | 整合多个反馈模块（冻结 LM 与外部工具），各模块聚焦特定错误类别，以实现更全面、独立的评估。 |
| N-CRITICS [795] | 实现批评者集成以评估初始输出。来自生成 LLM 与其他模型的汇集反馈指导精炼，直至满足停止准则。 |
| ISR-LLM [1383] | 通过将自然语言翻译为形式化规约、创建初始计划，再用验证器系统精炼，改进基于 LLM 的规划。 |
| SELF [710] | 以有限示例教授 LLM 元技能（自我反馈、自我精炼），随后模型通过生成并过滤自身训练数据持续自我演化。 |
| ProMiSe [892] | 针对较小 LM 的自我精炼，采用原则引导的迭代精炼，结合代理指标阈值与少样本精炼及拒绝采样。 |
| A2R [583] | 通过基于度量的迭代反馈学习增强 LLM，利用跨多维度（如正确性）的显式评估生成反馈并精炼输出。 |
| Experience Refinement [863] | 使 LLM 智能体在任务执行中通过从近期（相继）或全部先前（累积）经验中学习来精炼经验，优先高质量经验。 |
| I-SHEEP [660] | 允许 LLM 通过生成、评估、过滤并在高质量合成数据集上训练，从零开始持续自我对齐，无需外部指导。 |
| CaP [1280] | 使用外部工具精炼思维链(CoT)响应，应对模型陷入不可纠正推理循环的局限。 |
| Agent-R [1286] | 使语言智能体通过迭代自我训练“即时”反思，利用蒙特卡洛树搜索（MCTS）构造纠正错误路径的训练数据。 |
| GenDiE [616] | 以句子级优化增强上下文忠实性，结合生成与判别训练，赋予 LLM 自我生成与自我评分能力。 |
| Self-Developing [472] | 使 LLM 通过将改进算法生成为代码、评估它们并用 DPO 递归改进，自主发现、实现并精炼自身的改进算法。 |
| SR-NLE [1130] | 通过使用自我反馈与特征归因的迭代批判与精炼过程，改进事后自然语言解释的忠实性。 |
表 2： 大语言模型中的自我精炼方法及其关键特征。
4.1.2. 外部知识检索（External Knowledge Retrieval）
外部知识检索是上下文检索的关键组件，通过动态访问包括数据库、知识图谱与文档集合在内的外部信息源，应对参数化知识的基本局限。
检索增强生成基础。 RAG 将存储于模型参数中的参数化知识与从外部源检索的非参数化信息相结合，在保持参数效率的同时实现对当前、领域特定知识的访问 [597, 315, 257]。FlashRAG 提供 RAG 系统的全面评估与模块化实现，而 KRAGEN 与 ComposeRAG 等框架则展示了在多样基准测试上具有实质性性能提升的先进检索策略 [506, 755, 1168]。
Self-RAG 引入自适应检索机制：模型动态决定何时检索信息，并生成特殊 token 以控制检索时机与质量评估 [41]。先进实现包括用于层次化文档处理的 RAPTOR、受记忆启发的检索架构 HippoRAG，以及利用结构化知识表示改进信息访问的图增强 RAG 系统 [936, 370, 364]。
知识图谱集成与结构化检索。 知识图谱集成通过如 KAPING 等框架应对结构化信息检索：该框架基于语义相似性检索相关事实并将其前置到提示词中，无需模型训练 [48, 679]。KARPA 通过预规划、语义匹配与关系路径推理提供免训练的知识图谱适应，在知识图谱问答任务上达到最先进性能 [262]。
Think-on-Graph 使模型能够在知识图谱上顺序推理以定位相关三元组，开展探索以从外部数据库检索相关信息，同时生成多条推理路径 [1008, 726]。StructGPT 实现迭代的“先读后推理”方法，构造专门函数以从结构化数据源收集相关证据 [495]。
智能体式(agentic)与模块化检索系统。 智能体式 RAG 系统将检索视为动态操作，其中智能体充当智能调查员，分析内容并交叉参照信息 [654, 166, 973]。这些系统纳入复杂的规划与反思机制，需要整合任务分解、多计划选择与迭代精炼能力 [444, 1192]。
模块化 RAG 架构通过标准化接口与即插即用设计实现检索组件的灵活组合。图增强 RAG 系统利用结构化知识表示改进信息访问，而实时 RAG 实现则应对流式应用中的动态信息需求 [316, 1401]。
4.1.3. 动态上下文组装（Dynamic Context Assembly）
动态上下文组装代表将已获取信息组件编排为连贯、任务优化上下文的复杂过程，在尊重计算约束的同时最大化语言模型性能。
组装函数与编排机制。 组装函数 $A$ 涵盖基于模板的格式化、基于优先级的选择以及自适应组合策略，必须适应变化的任务需求、模型能力与资源约束 [708, 1142, 575]。当代编排机制在多智能体系统中管理智能体选择、上下文分发与交互流控制，通过用户输入处理、上下文分发以及基于能力评估的最优智能体选择实现有效协作 [902, 53, 175]。
先进编排框架纳入意图识别、上下文记忆维护与任务分派组件，以实现跨领域特定智能体的智能协调。Swarm Agent 框架利用实时输出引导工具调用，同时应对静态工具注册表与定制通信框架的局限 [814, 267, 250]。
多组件集成策略。 上下文装配必须应对跨模态集成挑战，纳入文本、结构化知识、时序序列以及外部工具接口等多样数据类型，同时保持连贯的语义关系 [535, 1230, 502]。言语化(verbalization)技术将知识图谱三元组、表格行与数据库记录等结构化数据转换为自然语言句子，从而在无需修改架构的情况下与现有语言系统无缝集成 [12, 788, 1072, 13]。
结构化数据的程序设计语言表示——尤其是知识图谱的 Python 实现与数据库的 SQL——在复杂推理任务中优于传统自然语言表示，因其能利用固有的结构属性 [1175]。多层次结构化方法依据语言关系将输入文本重组为分层结构；而结构化数据表示则利用现有大语言模型提取结构化信息，并将关键要素表示为图、表或关系模式 [687, 1134, 1334]。
自动化装配优化。 自动化提示工程(automatic prompt engineering)通过系统化的提示生成与精炼算法，解决人工优化的局限。Automatic Prompt Engineer (APE) 采用搜索算法发现最优提示；LM-BFF 则引入自动化流水线，将基于提示的微调与动态示例纳入相结合，在 NLP 任务上取得最高约 30% 的绝对提升 [311, 421, 596]。Promptbreeder 实现自指涉的进化系统，大语言模型同时改进任务提示以及支配这些改进的变异提示，类比自然选择 [279, 514]。
Self-Refine 通过多轮自我批判与修订实现输出的迭代改进；GPT-4 借此方法取得约 20% 的绝对性能提升 [741, 676]。多智能体协作框架模拟专业化团队动态，智能体承担分析师、编码者、测试者等不同角色，相较单智能体方法，在 Pass@1 指标上取得 29.9%–47.1% 的相对提升 [440, 1266]。
工具集成框架将思维链(CoT)推理与外部工具执行相结合，将中间推理步骤自动生成为可执行程序，并策略性地纳入外部数据。LangChain 为顺序处理链、智能体开发与网页浏览能力提供全面框架支持；Auto-GPT 与 Microsoft 的 AutoGen 等专用框架则以用户友好的接口促进复杂 AI 智能体的开发 [971, 1095, 25, 875]。
4.2. 上下文处理(Context Processing)
上下文处理聚焦于对已获取的上下文信息进行变换与优化，以最大化其对大语言模型的效用。该组件应对超长序列上下文的处理挑战，支持迭代式上下文自我精炼(Contextual Self-Refinement)与适应机制，并促进将多模态、关系型与结构化信息整合为连贯的上下文表示。
4.2.1. 长上下文处理(Long Context Processing)
超长序列上下文处理应对由 Transformer 自注意力 $O(n^{2})$ 复杂度引发的根本性计算挑战：随着序列长度增加，该复杂度形成显著瓶颈，并对实际应用产生实质性影响 [1067, 737, 299, 272, 420]。将 Mistral-7B 的输入从 4K 增至 128K token 需要约 122 倍的计算量增加；而预填充(prefilling)与解码阶段的内存约束亦造成巨大资源需求——Llama 3.1 8B 在每个 128K-token 请求上最高需约 16GB [1040, 1236, 429]。
面向长上下文的架构创新。 状态空间模型(State Space Models, SSMs)通过固定大小的隐状态保持线性计算复杂度与恒定内存需求；Mamba 等模型提供较传统 Transformer 更可扩展的高效递推计算机制 [1267, 351, 350]。LongNet 等膨胀注意力(dilated attention)方法随 token 距离增大而指数扩展注意力场，在保持 token 间对数依赖的同时实现线性计算复杂度，从而能够处理超过十亿 token 的序列 [220]。
Toeplitz Neural Networks (TNNs) 用相对位置编码的 Toeplitz 矩阵对序列建模，将时空复杂度降至对数线性，并支持从 512 个训练 token 外推至 14,000 个推理 token [876, 877]。线性注意力机制通过将自注意力表达为核特征映射的线性点积，将复杂度从 $O(N^{2})$ 降至 $O(N)$，在处理极长序列时可实现最高约 4000× 的加速 [528]。非注意力大语言模型等替代路径则借助递归记忆 Transformer 及其他架构创新突破二次方障碍 [553]。
位置插值与上下文扩展。 位置插值技术通过智能地重缩放位置索引（而非外推至未见位置），使模型能够处理超出原始上下文窗口限制的序列 [153]。神经切线核(Neural Tangent Kernel, NTK)方法为上下文扩展提供数学上有据的框架；YaRN 将 NTK 插值与线性插值及注意力分布校正相结合 [839, 477, 1029]。
LongRoPE 通过两阶段方法实现 2048K token 的上下文窗口：先将模型微调至 256K 长度，再进行位置插值以达到最大上下文长度 [222]。Position Sequence Tuning (PoSE) 通过组合多种位置插值策略，展示了高达 128K token 的显著序列长度扩展 [1387]。Self-Extend 技术采用双层注意力策略——分组注意力与邻近注意力——以捕获远距离与相邻 token 之间的依赖，从而在无需微调的情况下使大语言模型处理长上下文 [505]。
高效处理的优化技术。 Grouped-Query Attention (GQA) 将查询头划分成共享键/值头的组，在多查询注意力与多头注意力之间取得平衡，同时降低解码阶段的内存需求 [16, 1351]。FlashAttention 利用非对称的 GPU 记忆层次(memory hierarchy)，实现线性而非二次方的内存扩展；FlashAttention-2 通过减少非矩阵乘法运算并优化工作分配，提供约两倍速度 [200, 199]。
Ring Attention with Blockwise Transformers 通过跨多设备分布计算、利用分块计算并将通信与注意力计算重叠，从而支持极长序列 [682]。稀疏注意力技术包括 LongLoRA 中的 Shifted sparse attention (S²-Attn) 以及带 SF-Attn 的 SinkLoRA，在显著节省计算的同时达到全注意力困惑度改进的 92% [1313, 1226]。
Efficient Selective Attention (ESA) 提出通过将查询与键向量压缩为低维表示，在 token 级别选择关键信息，从而支持处理长达 256K token 的序列 [1092]。BigBird 将局部注意力与可关注整个序列的全局 token 相结合，并辅以随机连接，使高效处理的序列长度可达此前约 8× [1294]。
内存管理与上下文压缩。 内存管理策略包括 Rolling Buffer Cache 等技术：其维持固定注意力跨度，在 32K token 序列上将缓存内存使用降低约 8× [1351]。StreamingLLM 通过保留关键的“注意力汇(attention sink)”token 以及近期的 KV 缓存条目，在无需微调的情况下处理无限长序列；相对滑动窗口重计算，在长达 400 万 token 的序列上最高可加速 22.2× [1185]。
Infini-attention 将压缩记忆纳入普通注意力，在单个 Transformer 块中结合掩码局部注意力与长期线性注意力，从而以有界内存与计算处理无限长输入 [798]。Heavy Hitter Oracle (H₂O) 基于“少量 token 贡献大部分注意力价值”的观察，提出高效的 KV 缓存驱逐策略，吞吐量最高提升约 29×，延迟最高降低约 1.9× [1343]。
QwenLong-CPRS 等上下文压缩技术实现由自然语言指令引导的多粒度压缩动态上下文优化机制 [952]。Inf LLM 将远距离上下文存储于额外记忆单元，并采用高效机制检索与 token 相关的单元用于注意力计算，使在仅数千 token 序列上预训练的模型能够有效处理长达 1,024K token 的序列 [1184]。
4.2.2. 上下文自我精炼与适应(Contextual Self-Refinement and Adaptation)
上下文自我精炼(self-refinement)使大语言模型能够通过循环反馈机制——镜像人类修订过程——改进输出；其借助提示工程以对话式自我交互进行自我评估，有别于强化学习方法 [741, 924, 25, 1220]。
基础自我精炼框架。 Self-Refine 框架由同一模型兼任生成器、反馈提供者与精炼器，表明识别并修复错误往往比一次性产出完美初始解更容易 [741, 1322, 231]。Reflexion 将反思性文本保存在情节记忆缓冲中，供未来决策通过语言反馈使用 [964]；结构化引导被证明至关重要，因为过于简单的提示往往无法实现可靠的自我纠错 [678, 593]。
Multi-Aspect Feedback 整合冻结语言模型与专注特定错误类别的外部工具，以实现更全面、独立的评估 [805]。N-CRITICS 框架实施基于集成的评估：初始输出同时由生成该输出的大语言模型及其他模型评估，汇总反馈指导精炼，直至满足任务特定的停止准则 [795]。
A2R 框架对正确性与引用质量等多个维度进行显式评估，为各方面形成自然语言反馈并迭代精炼输出 [583]。ISR-LLM 通过将自然语言翻译为形式化规范、生成初始规划，再由验证器系统精炼，从而改进基于大语言模型的规划 [1383]。
元学习与自主演化。 SELF 以有限示例教会大语言模型元技能（自我反馈、自我精炼），随后模型通过生成并筛选自身训练数据持续自我演化 [710]。自我奖励机制使模型通过迭代式自我评判自主改进：单一模型同时扮演执行者与评判者，最大化其自我赋予的奖励 [1172, 1287]。
Creator 框架扩展了这一范式，使大语言模型通过涵盖创建、决策、执行与识别的四模块过程创建并使用自身工具 [954, 862]。Self-Developing 框架代表最为自主的路径：大语言模型通过迭代循环以可执行代码生成算法候选，从而发现、实现并精炼其自身的改进算法 [472]。
上下文学习(in-context learning)本质上是一种元学习形式：模型在预训练阶段习得可跨多样任务泛化的优化策略，从而在推理时快速适应新挑战 [183, 1174]。元上下文学习(meta-in-context learning)表明，上下文学习能力可通过上下文学习本身递归改进，自适应地重塑对期望任务的先验，并修改上下文学习策略 [181]。
记忆增强的适应框架。 记忆增强是实现元学习的有力途径，例如 Memory of Amortized Contexts 框架：其利用特征提取与记忆增强，将新文档信息压缩为紧凑调制量并存入记忆库 [1019]。Context-aware Meta-learned Loss Scaling 通过元训练小型自回归模型，在在线微调期间为每个 token 动态重加权语言建模损失，以应对过时知识挑战 [436]。
Decision-Pretrained Transformers 展示了如何训练 Transformer 执行上下文强化学习，通过超出预训练分布的泛化来求解此前未见的 RL 问题 [1021, 588]。基于上下文的元强化学习方法通过对上下文编码器的直接监督增强性能，相较端到端训练提高样本效率 [1080]。
长思维链与高级推理。 长思维链(Long Chain-of-Thought)已成为一项重要演进：其以显著更长的推理轨迹支持充分的问题探索，并体现在 OpenAI-o1、DeepSeek-R1、QwQ 与 Gemini 2.0 Flash Thinking 等先进模型中 [148, 724, 1223]。LongCoT 的有效性似乎与上下文窗口容量相关；经验证据表明，更大的上下文窗口往往带来更强的推理表现 [1238]。
扩展推理使模型能够在解题过程中进行自我反思与纠错 [1344]。即便不增加新信息，单纯增加推理步长亦可借助测试时扩展(test-time scaling)在多个数据集上显著增强推理能力 [1355]。
针对冗长推理轨迹带来的计算低效，优化策略包括：经 best-of-N 采样自生成更短推理路径、Zero-Thinking 与 Less-Thinking 等自适应推理模式，以及在保持推理质量的同时减少 token 用量的显式紧凑 CoT 方法 [797, 1358, 703]。Auto Long-Short Reasoning 可按问题复杂度动态调整推理路径长度，帮助模型判断何时需要更长的链条 [721]。
4.2.3. 多模态上下文(Multimodal Context)
多模态大语言模型(Multimodal Large Language Models, MLLMs)将上下文工程扩展至文本之外，把视觉、音频与三维环境等多样数据模态整合为统一的上下文表示。这一扩展在模态融合、跨模态推理与长上下文处理方面引入新挑战，同时使依赖丰富多模态上下文理解的复杂应用成为可能。
多模态上下文集成
基础技术。 多模态 MLLM 在传统大语言模型之上整合视觉、音频与三维环境等多样模态数据 [105, 49, 965]。一种主要集成方式是将视觉输入转换为离散 token，并与文本 token 拼接，从而使大语言模型的生成过程以联合表示为条件 [1295]。这通常由视觉提示生成器(Visual Prompt Generators, VPGs)促成：其在图像–字幕对上训练，以将视觉特征映射到大语言模型的嵌入空间 [613]。主导架构范式是通过 Q-Former 或简单 MLP 等对齐模块，将 CLIP（视觉）或 CLAP（音频）等专用外部多模态编码器连接到大语言模型主干 [19, 86, 615, 1139]；该模块化设计允许独立更新编码器而无需重训整个模型 [624]。
高级集成策略。 更精深的方法支持更深的模态融合。跨模态注意力机制直接在大语言模型嵌入空间内学习文本与视觉 token 之间的细粒度依赖，从而增强图像编辑等任务的语义理解 [570, 909, 102]。为管理冗长输入，分层设计分阶段处理各模态以保证可扩展性 [158]；“浏览–聚焦(browse-and-concentrate)”范式则在输入大语言模型之前融合多幅图像的上下文，以克服孤立处理的局限 [1143]。部分研究绕过对纯文本大语言模型的适配，转而采用统一训练范式：从一开始即在多模态数据与文本语料上联合预训练，以缓解对齐挑战 [1391, 1233]。另有方法将文本作为通用语义空间，利用大语言模型的上下文学习提升跨多样模态组合的泛化 [1058]。对于视频，上下文集成技术涵盖从提示调优到基于适配器的方法，将视频内容转换为可供推理的序列 [1088]。这些模型的发展往往受制于对海量高质量多模态数据与显著算力资源的需求 [1304, 615, 215]。
多模态上下文处理的核心挑战
模态偏置与推理缺陷。 MLLM 发展中的首要障碍是模态偏置(modality bias)：模型偏好文本输入，依赖已学语言模式而非整合的视觉或听觉信息，生成看似合理却未在多模态上接地(grounding)的回答 [1368, 24, 319, 1335]。训练方法会加剧该问题；例如，在简单图像字幕任务上训练的 VPG 仅学会抽取对字幕显著的特征，而忽略对更复杂、基于指令的任务至关重要的其他视觉细节，从根本上限制了深层多模态理解 [613, 510]。因此，MLLM 常在细粒度空间或时间推理上吃力，例如精确目标定位或理解视频中的细致事件序列 [1039, 965]，尤其在社交媒体等复杂领域——解读文本与图像交织以理解虚假信息或讽刺尤为困难 [511]。有效的多模态推理不仅要求理解各模态本身，还需推断其组合后的整体含义 [389]。雪上加霜的是，我们对 MLLM 内部机制的理解仍然有限：其内部运作大体上是黑箱，阻碍了更优架构的研发 [1283]。
高级上下文能力与未来方向
上下文学习与长上下文学习。 MLLM 的一项关键能力是上下文学习：模型无需更新权重，即可从提示中的多模态示例适应新任务 [1407, 1408, 557]。链接上下文学习(link-context learning, LCL)通过提供带显式因果链接的示范增强这一能力，从而改进泛化 [1020]。然而，上下文学习受固定上下文窗口制约——图像 token 占用大量空间，限制了多样本学习 [443]。性能亦对输入顺序敏感，且各模态的相对重要性因任务而异 [1028, 1206]。处理长多模态上下文（对视频分析等应用至关重要）仍是主要研究前沿 [1094]。创新包括面向视频的自适应分层 token 压缩 [1128]、可变视觉位置编码(V2PE) [1391]、用于对话记忆的 ContextQFormer 等专用模块 [595]，以及面向视频的动态、查询感知帧选择 [587]。MLLM 在延长交互中还表现出涌现的通信效率，该现象仍在研究之中 [442]。
新兴应用。 处理丰富多模态上下文的能力正在解锁新应用。MLLM 被用于预测性推理，例如从视觉场景预测人类活动 [1392]，并在多种多模态基准测试上展现出色的感知与认知能力 [294]。在视觉问答(VQA)中，上下文被用于更精确的回答，例如提示 MLLM 为图像生成自身的描述性文本上下文 [1356]，或通过检索增强生成(RAG)整合外部知识 [1001, 105]。其他应用包括基于感官输入规划数字动作 [611]、通过记忆增强的上下文理解强化手术决策支持 [422]，以及通过整合视觉信息与语音、音频线索实现细腻的视频理解 [648, 1202, 7]。研究者还将 MLLM 扩展至触觉信息、事件数据与图结构等新兴模态 [1368, 1031, 1222]。这些现实用例日益重要，推动了评估上下文理解的综合评估框架的发展 [1118]。这些进展使图像字幕与复杂多模态推理等纯文本模型此前无法实现的应用成为可能 [1182, 683, 139]。
4.2.4. 关系型与结构化上下文(Relational and Structured Context)
大语言模型在处理表格、数据库与知识图谱等关系型与结构化数据时面临根本性约束，源于基于文本的输入要求与顺序架构局限 [495, 47, 1145]。线性化往往无法保留复杂关系与结构属性；当信息分散于上下文各处时，性能会下降 [592, 591, 946]。
知识图谱嵌入与神经集成。 先进编码策略通过知识图谱嵌入将实体与关系转换为数值向量，以应对结构局限，从而在语言模型架构内高效处理 [12, 1259, 938, 1203]。图神经网络捕获实体间复杂关系，经 GraphFormers 等专用架构——在 Transformer 块旁嵌套 GNN 组件——促进跨知识图谱结构的多跳推理 [982, 408, 1230, 489]。
GraphToken 通过显式表示结构信息取得大幅改进，经参数高效的编码函数在图推理任务上最高提升约 73 个百分点 [842]。Heterformer 及其他混合 GNN–LM 架构在统一模型中执行上下文化文本编码与异构结构编码，以应对扩展此类集成系统的计算挑战 [502, 471, 757]。
| 方法 | 方法路径 | 性能 | 关键创新 |
|---|---|---|---|
| ODA [1009] | 观察驱动的智能体框架 | 12.87% 与 8.9% 的提升 | 带行动–反思的递归观察 |
| RAG-KG [1215] | 历史问题知识图谱构建 | 77.6% MRR，0.32 BLEU 提升 | 查询解析与子图检索 |
| KARPA [262] | 免训练知识图谱适配 | 知识图谱问答(KGQA) 达到当时最优 | 预规划关系路径 |
| Faithful Reasoning [726] | 规划–检索–推理框架 | N/A | 经关系路径的 LLM–KG 协同 |
表 3：面向大语言模型增强推理的知识图谱集成方法。
言语化与结构化数据表示。 言语化技术将知识图谱三元组、表格行与数据库记录等结构化数据转换为自然语言句子，从而在无需架构修改的情况下与现有语言系统无缝集成 [12, 788, 1072, 13]。多层次结构化方法依据语言关系将输入文本重组为分层结构；而结构化数据表示则利用现有大语言模型提取结构化信息，并将关键要素表示为图、表或关系模式 [687, 1134, 1334, 1043, 608]。
结构化数据的程序设计语言表示——尤其是知识图谱的 Python 实现与数据库的 SQL——在复杂推理任务中优于传统自然语言表示，因其能利用固有的结构属性 [1175]。使用结构化矩阵表示的资源高效方法，为在结构化数据任务上降低参数量同时保持性能提供了有前景的方向 [347]。
集成框架与协同方法。 知识图谱与语言模型的集成遵循不同范式，各具实现策略与性能权衡 [823, 1149]。预训练集成方法如 K-BERT 在训练期间注入知识图谱三元组以内化事实知识；而推理时方法则可在无需完整重训模型的情况下实现实时知识访问 [696, 1246, 718]。
知识图谱增强的大语言模型通过检索式增强方法纳入结构化知识以改进事实接地，例如 KAPING：其基于语义相似度检索相关事实并前置到提示中，无需模型训练 [48, 679, 597]。更精深的实现通过适配器模块与跨注意力机制，将知识图谱导出的表示直接嵌入模型潜在空间；Text2Graph 映射器则在输入文本与知识图谱嵌入空间之间提供链接 [132, 1074, 432]。
协同方法构建统一系统，使两类技术扮演同等重要角色，通过由数据与知识驱动的双向推理应对根本局限 [823, 859, 1120]。GreaseLM 促进所有模型层上的深度交互，使语言上下文表示由结构化世界知识接地，同时语言细微差别亦可信息化图表示 [1330]。
QA-GNN 通过联合图构建与基于图的消息传递，实现问答上下文与知识图谱之间的双向注意力(attention)机制及相互表示更新 [1259, 982]。
应用与性能增强。 结构化数据集成在多个维度显著增强大语言模型能力：知识图谱提供结构化信息，通过将回答锚定于可验证事实来减少幻觉，并借助清晰界定的信息来源提升事实准确性 [1010, 1352, 204, 571]。知识图谱通过提供结构化实体关系增强推理能力，支持复杂的多跳推理与逻辑推断；其丰富的层次化知识库显著提高推断的精度与可靠性 [1175, 212, 1026]。
真实世界应用在多个专业领域展现出实质性改进。医疗系统通过检索增强生成(Retrieval-Augmented Generation, RAG)框架将结构化医学知识与上下文理解相结合，以改进疾病进展建模与临床决策 [848, 589]。科研平台将发现组织为结构化知识以支持假设生成与研究空白识别；商业分析系统则在基于规则的精确性与人工智能模式识别之间取得平衡，以产生更具可操作性的洞见 [1336, 1070]。
问答系统受益于面向结构化数据源的自然语言接口；这种集成可构建更稳健的系统，能够处理多模态查询并提供个性化回答，从而突破静态知识库的局限 [1326, 1125, 922, 1215]。研究表明，相较于非结构化记忆方法，结构化知识表示可在公开数据集上将摘要性能分别提升 40% 与 14%；而链式关键(Chain-of-Key)策略通过动态结构化记忆更新可带来额外性能增益 [465]。
| 方法 | 数据类型 | 集成方法 | 关键创新 | 任务范围 |
|---|---|---|---|---|
| K-LAMP [48] | 知识图谱 | 基于检索的增强 | KAPING 框架 | 零样本问答 |
| Pan et al. [823] | 知识图谱 | 预训练与推理集成 | 协同化 LLM + KG | 多领域推理 |
| StructLM [1402] | 表格、图、数据库 | 指令微调 | 110 万示例数据集 | 18 个数据集、8 项 SKG 任务 |
| Shao et al. [946] | 表格、数据库、KG | 线性化方法 | 模式链接与语法预测 | Text-to-SQL 任务 |
表 4： 大语言模型中结构化数据集成的代表性方法。
4.3. 上下文管理（Context Management）
上下文管理(Context Management)关注 LLM 中上下文信息的高效组织、存储与利用。该组件应对有限上下文窗口所带来的根本性约束，发展复杂的记忆层次(Memory Hierarchies)与存储架构，并实现压缩技术，以在保持可访问性与连贯性的同时最大化信息密度。
4.3.1. 根本性约束
LLM 在上下文管理方面面临根本性约束，源于多数架构固有的有限上下文窗口规模：这显著降低模型在需要深入理解长文档的任务上的效能，并带来沉重的计算开销，从而阻碍需要快速响应与高吞吐量的应用 [1082]。尽管扩展上下文窗口使模型能够处理整篇文档并捕获更长程依赖，但传统 Transformer 架构的计算复杂度随序列长度呈二次增长，使处理极长文本的代价过高 [1007]。虽有 LongNet 等创新方法将该复杂度降至线性，但在窗口大小与泛化能力之间取得平衡仍具挑战 [1007, 220]。
实证证据揭示了“中间迷失”(lost-in-the-middle)现象：LLM 难以访问长上下文中间部分的信息，而当相关信息位于输入开头或结尾时表现显著更好 [128, 691, 654]。这种位置偏置严重损害扩展思维链(chain-of-thought)推理任务的表现——其中关键的早期结果易被遗忘，性能相较无先验上下文时可急剧下降多达 73% [128, 1147, 381]。
LLM 本质上独立处理每次交互，缺乏跨序列交换维持状态的原生机制以及稳健的自我验证机制；这些约束可追溯至哥德尔不完备定理所揭示的根本局限 [128, 372]。这种根本的无状态性需要显式管理系统来维持连贯的操作序列，并确保稳健的故障恢复机制 [128]。上下文管理面临彼此对立的挑战：上下文窗口溢出——模型因超出窗口限制而“遗忘”先前上下文；以及上下文坍塌(context collapse)——扩大的上下文窗口或对话记忆使模型无法区分不同对话上下文 [993]。研究表明，思维链提示(prompt)所声称的益处并非源于真正的算法学习，而是依赖于问题特定的提示，且随问题复杂度增加而衰减 [992]。长上下文处理的计算开销还带来管理键值(key-value)缓存的额外挑战：缓存随输入长度大幅增长，在延迟与准确性两方面均形成瓶颈；多轮与纵向交互挑战进一步复杂化上下文管理——有限的有效上下文阻碍纵向知识积累，而多样本提示的 token 需求挤压系统与用户输入的可用空间并减慢推理速度 [919, 725, 393]。
4.3.2. 记忆层次与存储架构
现代 LLM 记忆架构采用复杂的层次化设计，并按方法论途径组织，以克服固定上下文窗口的局限。受操作系统启发的层次化记忆系统实现虚拟内存管理概念；MemGPT 是这一路径的典范，其系统在有限上下文窗口（主存）与外部存储之间分页信息，类似于传统操作系统 [819]。此类架构由主上下文（含系统指令、FIFO 消息队列与可写草稿区）以及通过显式函数调用访问的外部上下文组成；借助函数调用能力进行记忆管理，可实现自主分页决策 [837]。PagedAttention 受操作系统中虚拟内存与分页技术启发，用于管理 LLM 中的键值缓存记忆 [57]。
动态记忆组织基于认知原理实现创新系统：MemoryBank 运用艾宾浩斯遗忘曲线理论，根据时间与重要性动态调整记忆强度 [1211, 1372]。ReadAgent 采用情节分页以切分内容、记忆摘要以生成简洁表示，以及交互式查找以检索信息 [1211]。压缩–检索器(compressor-retriever)架构通过使用基座模型前向函数压缩与检索上下文，支持终身上下文管理，并保证端到端可微性 [1245]。
架构适配通过内部修改增强模型记忆能力，包括增强的注意力机制、改进的键值缓存机制以及修改的位置编码 [164, 1362]。知识组织方法将记忆结构化为互联语义网络，以支持自适应管理与灵活检索；而以检索机制为导向的方法则将语义检索与记忆遗忘机制相结合 [521, 1372, 450]。
系统配置通过组织方式在效率与可扩展性之间取得平衡：集中式系统能高效协调任务，但随主题增加在可扩展性上吃力，易导致上下文溢出；去中心化系统可减少上下文溢出，但因智能体间查询而增加响应时间 [400]。混合方法在共享知识与专业化处理之间取得平衡，以实现半自主运行，应对在计算效率与上下文保真度之间取舍的挑战，并缓解记忆饱和——过度存储过往交互会导致检索低效 [164, 400]。上下文管理器组件(Context Manager Components)为大语言模型提供快照创建、中间生成状态恢复以及整体上下文窗口管理等基础能力 [763]。
4.3.3. 上下文压缩（Context Compression）
上下文压缩(Context Compression)技术通过降低计算与记忆负担、同时保留关键信息，使 LLM 能够高效处理更长上下文。基于自编码器的压缩通过上下文内自编码器(In-context Autoencoder, ICAE)实现显著的上下文缩减：ICAE 将长上下文压缩为紧凑记忆槽，LLM 可直接以其为条件，实现 4× 上下文压缩，并在推理时显著提升处理扩展上下文的能力，同时改善延迟与记忆占用 [321]。循环上下文压缩(Recurrent Context Compression, RCC)在受限存储空间内高效扩展上下文窗口长度，并通过指令重建技术应对指令与上下文均被压缩时模型响应变差的挑战 [447]。
记忆增强方法通过基于 kNN 的记忆缓存增强上下文管理：该缓存存储过往输入的键值对以供后续查找，借助基于检索的机制提升语言建模能力 [397]。对比学习方法提高记忆检索准确性；旁路网络在无需微调 LLM 的情况下缓解记忆陈旧；而巩固表示方法则动态更新过往 token 表示，使上下文窗口可任意扩大而不受固定记忆槽限制 [397]。
层次化缓存系统实现复杂的多层方案：激活再填充(Activation Refilling, ACRE)采用双层 KV 缓存——第 1 层缓存紧凑捕获全局信息，第 2 层缓存提供细致局部信息，并根据查询从 L2 动态再填充 L1 中的相关条目，以融合宏观理解与具体细节 [865]。Infinite-LLM 通过 DistAttention 将注意力计算分布到 GPU 集群、跨实例借用记忆的责任机制以及全局规划协调，应对动态上下文长度管理 [943]。KCache 通过将 K Cache 存放于高带宽内存、将 V Cache 保留在 CPU 内存，并基于注意力计算选择性复制关键信息，以优化推理 [943]。
多智能体分布式处理是一种新兴路径：使用基于 LLM 的多智能体方法以分布式方式处理海量输入，应对在处理大量外部知识时知识同步与推理过程中的核心瓶颈 [705]。对真实世界键值缓存访问模式的分析揭示，RAG 与智能体等工作负载具有高缓存可复用性，凸显了高效分布式缓存系统与优化元数据管理的必要性，以减少冗余并提升速度 [1399]。这些压缩技术可与其他长上下文建模方法相结合，进一步增强 LLM 高效处理与利用扩展上下文的能力，同时降低计算开销并保持信息完整性 [321]。
| 方法 | 策略 | 效率 | 准确性 | 长度管理 | 可扩展性 |
|---|---|---|---|---|---|
| O1-Pruner [724] | RL 微调 | N/A | +准确率，-开销 | 自动剪枝 | +效率 |
| InftyThink [1223] | 迭代 + 摘要 | 复杂度降低 | +3–13% | 迭代控制 | 可扩展 |
| Long-CoT Survey [148] | 长 CoT + 推理 | +效率框架 | +复杂领域 | 深度探索 | 测试时扩展 |
| PREMISE [1282] | 提示优化 + 诊断 | 梯度启发优化 | 维持/+准确率 | −87.5% tokens | 性能维持 |
| Prune-on-Logic [727] | 结构感知剪枝 | 选择性剪枝 | +准确性 | 选择性框架 | 基于逻辑的优化 |
表 5： 大语言模型中长链推理方法及其特征。O1-Pruner 使用类强化学习微调缩短推理链同时保持准确性。InftyThink 采用带中间摘要的迭代推理以降低计算复杂度。Long-CoT Survey 探讨长思维链特征如何通过效率改进与增强的知识框架提升推理能力。PREMISE 以轨迹级诊断结合梯度启发优化对提示进行优化，实现 87.5% 的 token 缩减。Prune-on-Logic 通过选择性移除低效用推理步骤，对逻辑图进行结构感知剪枝。
4.3.4. 应用
有效的上下文管理使 LLM 的能力超越简单问答，支撑跨多领域、依托全面上下文理解的复杂应用。文档处理与分析能力使 LLM 能够处理整篇文档或理解完整文章而非片段，通过全面理解输入材料给出上下文相关的回答；这对基因序列、法律文书与技术文献等本身冗长的序列数据尤为重要，因为在广延内容上保持连贯性至关重要 [1007]。
由上下文管理技术促成的扩展推理能力支持需要在扩展序列中维持并基于中间结果进行构建的复杂推理。通过捕获更长程依赖，这些系统支持多步问题求解——其中后续推理依赖先前的计算或演绎——从而赋能需要广泛上下文感知的复杂决策支持系统与科研助手等应用 [1007, 164]。
协作与多智能体系统受益于多轮对话或顺序任务中的有效上下文管理：在此类场景中，维持一致状态并在协作模型间同步内部信息至关重要 [157]。这些能力支撑分布式任务处理、协作内容创作以及多智能体问题求解等应用，要求跨多次交互保持上下文连贯性 [157]。
增强的会话界面依托稳健的上下文管理无缝处理长对话而不丢失线索连贯性，从而实现更自然、可持续的对话，更接近人类交谈 [891]。面向任务的 LLM 系统受益于结构化上下文管理方法：滑动窗口存储实现最小上下文管理系统，将提示与回答永久追加到上下文存储；检索增强生成系统则为 LLM 补充对动态外部信息源的访问 [216, 934]。这些能力支撑个性化虚拟助手、长期辅导系统以及在扩展交互中保持连续性的治疗性会话智能体等应用 [891]。
记忆增强应用实现使 LLM 能够持久存储、管理并动态检索相关上下文信息的策略，支撑需要随时间积累知识的应用：通过持续交互构建个性化用户模型、在扩展交互中实施有效知识管理，以及支持依赖历史上下文的长期规划场景 [164]。诸如上下文感知智能记忆(Contextually-Aware Intelligent Memory, CAIM)等先进记忆框架，通过融入认知 AI 原则的模块增强长期交互：这些模块支持用户特定信息的存储与检索，并支持基于上下文与时间的相关性过滤 [1152]。面向 LLM 智能体的记忆管理纳入类似人类记忆再巩固的过程，包括去重、合并与冲突解决；反思记忆管理(Reflective Memory Management)等方法结合前瞻与回顾反思，以实现动态摘要与检索优化 [1176, 386]。基于案例的推理系统为 LLM 智能体记忆提供理论基础：其架构组件支持认知集成，持久上下文存储技术则实现缓存策略，以更快地供给必要上下文 [387, 385]。这些收益不仅在于处理更长文本，更从根本上通过提升理解力、生成更相关回答以及在扩展互动中保持更大连续性来增强 LLM 交互质量，从而显著拓展 LLM 的效用，并化解受限于受限上下文窗口的局限 [891]。
5. 系统实现（System Implementations）
在上下文工程(Context Engineering)的基础组件(Foundational Components)之上，本节考察将这些组件整合为实用智能架构的复杂系统实现(System Implementations)。这些实现体现了从理论框架到可部署系统的演进，后者运用上下文工程原则。我们介绍四大类系统实现。RAG 系统通过模块化架构与图增强方法展示外部知识集成。记忆系统通过支持长期学习的复杂记忆架构展示持久上下文管理。工具集成推理通过函数调用与环境交互将语言模型转化为世界交互者。多智能体系统通过通信协议与编排机制呈现协调方法。每类实现均建立在基础组件之上，同时应对上下文利用中的特定挑战，展示理论原则如何转化为实用系统。
5.1. 检索增强生成（Retrieval-Augmented Generation）
检索增强生成(Retrieval-Augmented Generation)通过将外部知识源与语言模型生成相集成，弥合参数化知识与动态信息访问之间的鸿沟。该实现使模型能够通过模块化架构、智能体式框架与图增强方法访问当前的、领域特定的信息，从而超越静态训练数据。
5.1.1. 模块化RAG架构
模块化RAG(Modular RAG)从线性检索–生成架构转向具有灵活组件交互的可重构框架 [315, 1140, 597]。不同于朴素 RAG(Naive RAG)与高级 RAG(Advanced RAG)的查询改写，模块化 RAG 引入层次化架构：顶层 RAG 阶段、中层子模块与底层操作单元 [316, 736]。这通过路由、调度与融合机制超越线性结构，实现动态重构 [316]。
形式化表示 $\mathrm{RAG} = {R, G}$ 通过复杂的模块编排运作，支持改写–检索–阅读(Rewrite-Retrieve-Read)模型与生成–阅读(Generate-Read)方法，并纳入自适应搜索模块、用于多查询处理的 RAGFusion、用于最优数据源选择的路由模块，以及兼顾检索准确性与上下文相关性的混合检索策略 [315, 497, 916, 1053, 888, 95]。
图 4： 检索增强生成框架：RAG 系统架构概览，包括模块化RAG、智能体式RAG系统以及用于外部上下文集成的图增强RAG方法。（原文图见 PDF 第 27 页）
当代框架在检索准确性与可信度方面展现显著改进 [1382]。FlashRAG 提供含 5 个核心模块与 16 个子组件的模块化工具包，支持独立调整与流水线组合 [506]。KRAGEN 通过将知识图谱与向量数据库集成增强生物医学问题求解，利用面向生物医学知识图谱优化的提示生成，以应对复杂推理中的幻觉 [401, 755, 981]。ComposeRAG 实现用于问题分解与查询改写的原子模块，并纳入自我反思机制以进行迭代精炼 [1168]。这种模块性便于与微调及强化学习集成，支持面向特定应用的定制化，以及覆盖多样化 NLP 任务的综合工具包 [316, 920, 4]。
5.1.2. 智能体式RAG系统（Agentic RAG）
智能体式RAG(Agentic RAG)将自主 AI 智能体嵌入 RAG 流水线，实现由持续推理引导的动态、上下文敏感操作 [973, 281]。这些系统借助反思、规划、工具使用与多智能体协作，动态管理检索策略，并使工作流适配复杂任务需求 [973]。RAG 与智能体工作流通过对齐而关联：查询改写对应语义理解，检索阶段对应规划与执行 [628]。
基于 LLM 的自主智能体通过多模态感知、工具利用与外部记忆集成扩展基础语言模型能力 [1169, 1099, 939, 849]。外部长期记忆作为知识数据存储，使智能体能够在较长时期内纳入并访问信息 [1169, 386]。不同于静态方法，智能体式 RAG 将检索视为动态操作，智能体充当智能调查员，分析内容并交叉核验信息 [654, 166]。
实现范式包括无需额外训练的基于提示的方法，以及通过强化学习优化模型以进行策略性工具调用的基于训练的方法 [654, 1327, 973]。先进系统使 LLM 智能体能够在单一工作流中查询向量数据库、访问 SQL 数据库或调用 API；方法论进展聚焦于推理能力、工具集成、记忆机制以及面向自主决策的指令微调 [709, 6]。
核心能力包括通过任务分解、多计划选择与记忆增强规划策略实现的推理与规划组件，使智能体能够拆解复杂任务并选择合适策略 [444, 445]。PlanRAG 通过先规划再检索的方法改进决策，使智能体能够评估多信息源并优化检索策略；SLA 管理框架则面向可重构多智能体架构 [166, 467]。工具利用使系统能够调用搜索引擎、计算器与 API 等多样化资源；ReAct 与 Reflexion 等框架展示了将推理与行动交错如何增强适应性 [166, 1169, 964]。记忆机制提供外部长期存储，而自适应检索策略则支持对复杂度与上下文的自主分析 [166, 1137]。
自我反思与适应机制使智能体式 RAG 系统能够通过基于先前交互结果精炼操作的迭代反馈环，在动态环境中运行 [1192, 692]。MemoryBank 等先进记忆系统实现受艾宾浩斯遗忘曲线启发的更新机制，增强智能体从过往交互中检索并应用所学的能力 [1372, 169]。CDF-RAG 采用闭环过程，将因果图检索与强化学习驱动的查询精炼及幻觉纠正相结合 [537]。Self-RAG 训练按需检索段落、并对其检索与生成进行反思的模型，在推理时使用反思 token 控制行为 [243, 41]。
5.1.3. 图增强RAG（Graph-Enhanced RAG）
基于图的检索增强生成从面向文档的方法转向捕获实体关系、领域层次与语义连接的结构化知识表示 [120, 1363, 364, 1401]。这使得能够提取向语言模型提供相关信息的特定推理路径，并通过结构化路径导航支持多跳推理 [120]。图结构借助互联性增强上下文感知检索与逻辑连贯性，从而最小化上下文漂移与幻觉 [518, 812]。
知识图谱作为基础表示，以结构化格式封装实体及其相互关系，支持高效查询与语义关系捕获 [166, 1066]。基于图的知识表示可划分为：以图作为知识载体的基于知识的 GraphRAG、以图作为索引工具的基于索引的 GraphRAG，以及结合两者的混合 GraphRAG [1208]。复杂实现包括带社区检测的 GraphRAG 层次化索引、将文档组织为三层层次结构的 PIKE 多层次异构知识图谱，以及 EMG-RAG 的可编辑记忆图(Editable Memory Graph)架构 [317]。
图神经网络通过应对处理结构化知识方面的局限增强 RAG 系统；GNN 擅长捕获实体关联并提升知识一致性 [232, 116]。GNN-RAG 实现采用轻量级架构以有效检索知识图谱元素，在与语言模型对接前改进图结构捕获 [1380, 166]。集成过程涵盖通过节点与边提取构建图、基于查询的检索，以及纳入检索信息的生成 [1380]。
多跳推理能力使基于图的系统能够跨多个相连知识图谱节点综合信息，从而促进需要互联事实整合的复杂查询求解 [1066, 170]。这些系统采用结构化表示，以非结构化文本无法实现的方式捕获实体间语义关系与领域层次 [1066, 170]。Hierarchical Lexical Graph 等先进框架在保留陈述来源的同时对主题聚类以支持灵活检索，并链接实体以进行基于图的遍历 [333]。GraphRAG、LightRAG 及其衍生系统实现双层检索、层次化索引与图增强策略，以支持稳健的多层次推理 [1183, 317]。
主流架构展现了图增强检索的多样化方法，其优化策略在检索有效性方面取得了显著提升 [106]。LightRAG 通过双层检索范式将图结构与向量表示相结合，提升了效率与内容质量 [416, 723]。HippoRAG 利用知识图谱上的个性化 PageRank，在多跳问答中取得了显著改进 [1096, 752, 370]。HyperGraphRAG 提出超图结构化表示，超越了二元关系 [723]。RAPTOR 提供层次化摘要树构建以实现递归上下文生成，而 PathRAG 则为基于图的检索引入了剪枝技术 [1359, 936, 134]。这些结构化方法能够实现带有显式实体连接的透明推理路径，降低噪声并改善语义理解，同时克服传统检索增强生成(RAG)的挑战 [1183, 518]。
5.1.4. 应用
实时 RAG 系统应对生产环境中的关键挑战，其中动态知识库需要持续更新与低延迟响应 [1349, 534]。核心挑战包括高效部署与处理流水线优化，现有框架缺乏即插即用方案，因而需要系统级优化 [1349]。流式数据的集成引入了额外复杂性，因为传统架构在信息频繁变化时准确率较差，且随着文档量增长效率下降 [520]。
动态检索机制通过在生成过程中持续更新策略，超越了静态方法：基于生成状态与已识别的知识缺口，实时调整目标与语义向量空间 [388]。当前在确定最优检索时机与查询构造方面的局限，通过思维链(Chain-of-Thought)推理、迭代检索过程、分解式提示，以及用于动态检索的 LLM 生成内容等方法加以应对，从而实现自适应信息选择；相关方法还扩展到通过反思标签增强生成质量的自适应控制机制 [1000, 536, 85, 539, 1248]。
低延迟检索方法利用基于图的技术，在速度–准确率优化方面展现出显著潜力，稠密段落检索技术则提供了基础性改进 [525]。LightRAG 的双层检索系统增强了信息发现能力，同时将图结构与向量表示相结合以实现高效的实体关系检索，在保持相关性的同时缩短响应时间 [364]。多阶段检索流水线通过基于图的重排序等技术优化计算效率，在降低存储需求的同时实现对当前信息的动态访问 [982]。
可扩展性方案采用分布式处理架构，结合高效的数据分区、查询优化，以及适应变化流条件的容错机制 [1048, 35]。通过转换后的重击者(heavy hitters)流式算法进行内存优化，可智能过滤无关文档并保持质量，这对内容频繁变化的场景尤为有价值 [520]。生产框架通过支持查询扩展等预检索过程以及压缩与选择等后检索精炼的模块化 RAG 架构展现效率提升，从而能够对各个组件进行微调 [1077]。
增量索引与动态知识更新确保系统无需完全重新训练即可适应新信息，这在网络安全与气候金融等快速演进的领域尤为关键 [836, 1064]。现代框架纳入动态知识检索方法，能够基于不断演进的输入与上下文信息持续调整策略，增强交互性与语义理解，同时提升跨领域集成的适用性 [388]。先进的基于智能体的方法在复杂环境中展现出精细的任务分配能力，例如需要实时决策的协同无人机操作，其应用还延伸到具身智能体的接地规划 [1324, 983]。动态检索增强生成框架如 DRAGON-AI 展示了面向本体生成的专门实现，结合文本与逻辑组件，并纳入自记忆机制以实现迭代改进 [1051]。这些进展标志着向在动态环境中无缝集成实时知识与灵活检索能力的显著演进。
5.2. 记忆系统(Memory Systems)
记忆系统(Memory Systems)通过实现持久化的信息存储、检索与利用机制，使大语言模型(LLM)能够超越无状态交互。这一实现将模型从模式匹配处理器转变为能够学习、适应，并在扩展交互中具备长期上下文理解能力的复杂智能体。
图 5：记忆系统框架：面向 LLM 超长上下文处理的记忆架构、记忆增强智能体与评估挑战概览。（原文图见 PDF 第 30 页）
5.2.1. 记忆架构(Memory Architectures)
记忆使复杂语言系统区别于模式匹配模型，使其能够跨自然语言任务进行信息处理、存储与利用 [1191, 1176, 300]。尽管在文本生成与多轮对话方面取得突破，LLM 仍面临相当大的记忆系统约束 [1191]。神经记忆机制受困于结构化信息存储不足，以及依赖近似向量相似度计算而非精确符号运算，这对多跳推理中的准确存储与检索构成挑战 [427]。这些局限是开发在复杂现实世界应用中有效运行的 AI 系统所面临的关键挑战 [550]。
记忆分类框架。 LLM 记忆系统可按多种分类框架加以组织。主要的时间分类将记忆划分为三类：感觉记忆(sensory memory，输入提示)、短期记忆(short-term memory，即时上下文处理)，以及长期记忆(long-term memory，外部数据库或专用结构) [943]。从持久性视角看，短期记忆包括仅存在于单次会话中的键值缓存与隐藏状态，而长期记忆涵盖基于文本的存储以及嵌入模型参数中的知识，可跨多个交互周期持久存在 [943, 824]。
基于实现方式的分类识别出参数记忆(parametric memory，编码于模型权重中的知识)、短暂激活记忆(ephemeral activation memory，受上下文限制的运行时状态)，以及通过检索增强生成(RAG)方法访问的明文记忆(plaintext memory) [643]。当前实现缺乏精细的生命周期管理与多模态集成，限制了长期知识演化。前馈网络层作为存储记忆的键值表，充当词语检索的“内部词库”，并形成类似人类联想记忆的机制 [524, 329, 330, 770, 470]。这些分类方案反映了构建与人类认知系统相平行的 LLM 记忆架构的尝试 [1176]。
短期记忆机制。 LLM 中的短期记忆通过上下文窗口运作，充当工作记忆(working memory)，维持对先前处理过的 token 的即时访问 [1291]。该功能通过存储 token 表示的键值缓存实现，但在会话终止时消失 [899]。架构变体展现出显著差异：基于 Transformer 的模型实现工作记忆系统，可在任意延迟下灵活检索单个 token 表示；而 LSTM 架构则维持更粗粒度、快速衰减的语义表示，并偏向最早的条目 [40]。
现代 LLM 短期记忆常表现为上下文学习，反映模型在上下文窗口内临时获取与处理信息的能力 [1189, 103]。这使得无需参数更新即可进行少样本学习与任务适应。研究识别出三种主要记忆配置：完整记忆（利用全部上下文历史）、有限记忆（使用上下文子集），以及无记忆操作（不使用历史上下文）[1052]。尽管进展已将上下文窗口扩展至数百万 token，LLM 在对扩展上下文进行有效推理方面仍存在困难，尤其当相关信息出现在中间位置时 [899, 691]。
长期记忆实现。 由于上下文窗口限制与灾难性遗忘，LLM 在维持长期记忆方面面临重大挑战 [114]。基于外部记忆的方法通过利用物理存储缓存历史信息来应对这些限制，从而无需在受限上下文窗口内保留全部信息即可检索相关历史 [688, 1372]。这些方法与关注降低自注意力计算成本以扩展序列长度的基于内部记忆的方法形成对比 [688, 291]。
长期记忆实现可归类为知识组织方法（将记忆结构化为相互连接的语义网络）、面向检索机制的方法（将语义检索与遗忘曲线机制相结合），以及架构驱动方法（实现带有显式读写操作的层次结构）[521, 1372, 450]。记忆存储表示可进一步划分为 token 级记忆（信息以结构化文本形式存储以便直接检索）与潜在空间记忆（利用高维向量进行抽象且紧凑的信息表示）[1225, 1133]。先进方法融入心理学原理，其中 MemoryBank 实现艾宾浩斯遗忘曲线理论，基于时间因素进行选择性记忆保留 [1372]；情感感知框架采用心境依存记忆理论 [450]；记忆机制则通过抽取脆弱性分析在性能优势与隐私关切之间取得平衡 [1049, 122, 123]。
记忆访问模式与结构。 LLM 展现出与人类认知过程具有显著相似性的特征性记忆访问模式，在回忆信息列表时表现出清晰的首因效应与近因效应 [483]。记忆检索通过顺序访问（按连续顺序检索内容）与随机访问（从任意点访问信息而无需处理前述内容）运作 [1397]。记忆持久性研究采用识别实验、回忆实验与保持实验，以量化信息可访问时长与检索条件 [816]；语义记忆与情景记忆(episodic memory)整合等认知心理学概念可提升 LLM 的信息综合能力 [244]。
记忆组织涵盖多样化结构方法，包括文本形式存储（完整与近期的智能体–环境交互、检索到的历史交互、外部知识）、知识表示结构（块、知识三元组、原子事实、摘要、混合方法）、带有库增强推理组件的层次系统，以及按任务、时间相关性或语义关系组织的功能模式 [1339, 1299, 1035]。核心记忆操作包括编码（将文本信息转换为潜在空间嵌入）、检索（基于语义相关性、重要性与近因性访问相关信息）、反思（提取更高层洞察）、摘要（压缩文本同时突出关键要点）、利用（整合记忆组件以形成统一输出）、遗忘（选择性丢弃信息）、截断（在 token 限制内进行格式化），以及判断（评估信息重要性以确定存储优先级）[1341]。这些结构在全面性、检索效率与计算需求之间提供不同权衡。
5.2.2. 记忆增强智能体(Memory-Enhanced Agents)
记忆系统从根本上将 LLM 从无状态模式处理器转变为能够在扩展交互中进行持久学习与适应的复杂智能体 [1268]。记忆增强智能体(Memory-Enhanced Agents)同时利用短期记忆（促进实时响应与即时上下文感知）与长期记忆（支持在更长时期内的深层理解与知识应用），以适应变化的环境、从经验中学习，并做出需要持久信息访问的知情决策 [1268]。
智能体架构集成。 当代 LLM 智能体采用类似于计算机内存层次结构的记忆系统：短期记忆作为上下文窗口内上下文理解的主存储，而长期记忆则作为扩展信息保留的持久存储 [776]。从面向对象视角看，AI 系统生成与个体用户相关的个人记忆，以及包含中间任务结果的系统记忆 [1176]。诸如 MemOS 等结构化框架将记忆划分为参数记忆（编码于模型权重中的知识）、激活记忆与明文记忆；参数记忆表示嵌入前馈与注意力层中的长期知识，支持零样本生成 [643]。
记忆集成框架已演进为通过精细架构应对 LLM 局限。自控记忆(Self-Controlled Memory, SCM)框架通过基于 LLM 的智能体主干、记忆流以及管理更新与利用的记忆控制器来增强长期记忆 [655]。REMEMBERER 框架为 LLM 配备经验记忆，以跨任务目标利用过往情节，通过言语强化与自我反思反馈机制，无需参数微调即可实现成败学习 [1308]。诸如 MemLLM 等先进系统实现结构化读写记忆模块，应对稀有事件记忆、信息更新与防止幻觉等挑战 [785]。利用 LLM 的自主智能体依赖四个基本组件——感知、记忆、规划与行动——协同工作以实现环境感知、交互回忆以及实时规划与执行 [620, 38]。
| 文本形式 | 参数形式 | |||||
|---|---|---|---|---|---|---|
| 模型 | 完整 | 近期 | 检索 | 外部 | 微调 | 编辑 |
| 核心记忆系统 | ||||||
| MemoryBank [1373] | × | × | ✓ | × | × | × |
| RET-LLM [784] | × | × | ✓ | × | × | × |
| ChatDB [427] | × | × | ✓ | × | × | × |
| TiM [689] | × | × | ✓ | × | × | × |
| Voyager [1086] | × | × | ✓ | × | × | × |
| MemGPT [820] | × | ✓ | ✓ | × | × | × |
| RecMind [1124] | ✓ | × | × | × | × | × |
| Retroformer [1258] | ✓ | × | × | ✓ | ✓ | × |
| ExpeL [1347] | ✓ | × | ✓ | ✓ | × | × |
| Synapse [1367] | × | × | ✓ | × | × | × |
| 基于智能体的系统 | ||||||
| ChatDev [861] | ✓ | × | × | × | × | × |
| InteRecAgent [456] | × | ✓ | ✓ | ✓ | × | × |
| TPTU [917, 560] | ✓ | × | × | ✓ | × | × |
| MetaGPT [413] | ✓ | × | × | × | × | × |
| S3 [305] | × | × | ✓ | × | × | × |
| Mem0 [173] | × | × | ✓ | × | × | × |
| 先进记忆架构 | ||||||
| Larimar [202] | × | ✓ | ✓ | × | × | ✓ |
| EM-LLM [290] | × | ✓ | ✓ | × | × | × |
| Controllable Working Memory [603] | ✓ | ✓ | ✓ | × | ✓ | × |
| Working Memory Hub [359] | ✓ | ✓ | ✓ | ✓ | × | × |
| 近期与新兴系统 | ||||||
| LLM-based Opinion Dynamics [179] | × | × | ✓ | × | × | × |
| Memory Sandbox [462] | × | × | ✓ | × | × | ✓ |
| A-MEM [1212] | × | × | ✓ | × | × | ✓ |
| MemEngine [1341] | × | × | ✓ | ✓ | × | × |
| HIAGENT [433] | × | ✓ | ✓ | × | × | × |
| MemInsight [925] | × | × | ✓ | ✓ | × | × |
| Memory Sharing (MS) [306] | × | × | ✓ | ✓ | × | × |
| MemoRAG [866] | ✓ | × | ✓ | ✓ | ✓ | × |
| Echo [700] | ✓ | ✓ | ✓ | ✓ | ✓ | × |
表 6：扩展自 [1339]：记忆实现模式。✓= 已采用，×= 未采用
现实世界应用。 记忆增强 LLM 智能体已在多样化应用领域展现变革性影响。在对话式 AI 中，记忆系统通过回忆过往体验与用户偏好，交付个性化、上下文感知的响应，从而实现更自然、类人的交互。商业实现包括 Charlie Mnemonic（结合使用 GPT-4 的长期、短期与情景记忆）、Google Gemini（利用长期记忆在 Google 生态系统中提供个性化体验），以及 ChatGPT Memory（跨会话记住对话）[584]。用户模拟应用采用由 LLM 驱动的对话智能体模仿人类行为，以进行成本效益高的对话系统评估，并可灵活适应开放域对话、任务导向交互与对话式推荐 [208]；诸如 Memory Sandbox 等系统通过数据对象操作使用户能够控制对话记忆 [461]。
任务导向智能体利用记忆以最少的人工干预执行复杂自主操作，将 LLM 作为控制器，并通过多模态感知、工具利用与外部记忆加以扩展 [1169]。应用涵盖推荐系统（RecMind 通过规划与外部知识提供个性化推荐，InteRecAgent 将推荐模型作为工具与 LLM 结合使用）、自动驾驶（DiLu 通过推理、反思与记忆注入类人知识）、科学研究（ChemCrow 自动化化学合成设计与执行），以及社会仿真（生成式智能体通过记忆存储与综合展现可信行为）[1027, 653, 92, 831]。主动对话智能体应对战略对话场景中的挑战，这些场景需要目标导向的对话引导，通过基于提示的策略规划方法以及基于对话历史的 AI 反馈生成加以实现 [208, 207]。
个性化助手应用利用记忆与用户维持连贯的长期关系，记忆组件作为结构化仓储，存储包括用户偏好与历史交互在内的上下文相关信息 [444]。领域特定实现包括采用记忆协调进行医疗交互的医疗助手 [1325, 1316]、利用外部知识库的推荐智能体 [1325, 1302]、通过记忆支持的进度跟踪提供上下文感知支持的教育智能体 [653]，以及像 MARK 这样通过用户偏好记忆增强个性化 AI 助手的专门框架 [303]。
记忆技术与集成方法。 记忆技术演进通过 RAG 应对根本性的上下文窗口限制：RAG 结合参数与非参数记忆，使用预训练的 seq2seq 模型与稠密向量索引进行语言生成 [1218, 597]。该方法无需重新训练即可访问参数存储之外的信息，显著扩展知识能力。包括向量数据库与检索增强生成在内的先进记忆机制，能够实现海量信息存储与快速相关数据访问，并纳入短期上下文记忆与长期外部存储 [38, 371, 1193, 513]。
非参数方法在保持 LLM 参数冻结的同时，利用 RAG 等外部资源丰富任务上下文 [942]。诸如 Reflexion 等系统通过情景记忆缓冲区中的自我反思反馈实现言语强化，而 REMEMBERER 则纳入持久经验记忆，从而能够从过往成败中学习。诸如 MemoryBank 等先进架构通过整合先前交互信息，支持记忆检索、通过更新实现持续演化，以及个性适应 [1211, 1372]。
专门记忆架构通过精细的组织与检索机制应对特定智能体需求。早期系统需要预定义的存储结构与检索时机，而像 Mem0 这样的较新系统则遵循 RAG 原则纳入图数据库，以实现更有效的记忆组织与基于相关性的检索 [1211]。包括 OpenAI ChatGPT Memory、Apple Personal Context、mem0 与 MemoryScope 在内的商业与开源实现，表明记忆系统在增强个性化能力方面已被广泛采用 [1176]。工具增强范式验证了在复杂任务分解中的有效性，同时利用与世界交互的工具；记忆增强智能体正成为现代 AI 系统的核心，通过自然语言整合规划、工具使用、记忆与多步推理来执行复杂任务 [251, 360, 1099, 34]。
5.2.3. 评估与挑战
记忆评估框架已成为系统评估 LLM 智能体在多维度能力方面的关键组件，反映了智能系统中记忆的多面性。这些全面评估方法揭示了重大挑战，同时指向有望解锁记忆增强智能体新能力的研究方向。
评估框架与指标。 当代记忆评估采用专门指标，超越传统 NLP 性能指标，以捕捉记忆功能的细微方面 [1340]。有效性指标通过准确率（基于历史消息的响应正确性）与 recall@5 指标（在前 5 个结果中检索到相关消息的百分比），聚焦事实信息的存储与利用。效率指标通过响应时间（信息检索与利用所需时长）与适应时间（新信息存储所需时段）考察时间维度 [1340]。
诸如 LongMemEval 等大规模基准测试评估五项基本长期记忆能力：信息抽取、时间推理、多会话推理、知识更新与弃答，通过 500 个精心选取的问题，表明商业助手在长时间交互中准确率下降约 30%；自动化记忆评估框架则促进超越通关密钥搜索方法的全面评估 [1180]。专门框架通过评估时间定位体验的基准测试面向情景记忆；研究表明，包括 GPT-4、Claude 变体与 Llama 3.1 在内的前沿模型，即使在相对较短的上下文中，面对涉及相互关联事件或复杂时空关联的情景记忆挑战时仍遇到困难 [463]。当代 LLM 基准测试主要集中于评估模型对事实信息与语义关系的保持，却实质性地忽视了情景记忆评估——即将记忆与时间与空间发生细节进行情境化的能力 [847]。
任务特定评估涵盖长上下文段落检索（在扩展上下文中定位特定段落）、长上下文摘要（形成全面理解以生成简洁摘要）、NarrativeQA（基于冗长叙事回答问题），以及像 MADail-Bench 这样在对话语境中评估被动与主动记忆回忆的专门基准测试，其新颖维度包括记忆注入、情感支持能力与亲密度评估 [1339, 1390, 556, 390]。其他任务特定框架包括用于会议摘要的 QMSum、用于阅读理解的 QuALITY、需要时空记忆的基于对话问答的 DialSim，以及使用两阶段过程评估物理环境任务中记忆利用的个性化具身智能体评估框架 MEMENTO [1390, 572]。
当前局限与挑战。 记忆评估面临显著限制能力有效评估的挑战。根本局限包括缺乏评估记忆性能的一致、严格方法论，尤其是在训练数据之外的泛化方面 [288]。缺乏专门为长期记忆评估设计的标准化基准测试是另一重大障碍，现有框架往往无法捕捉类人智能所需记忆能力的全谱 [1079]。
架构约束显著增加了评估难度，因为大多数当代基于 LLM 的智能体以根本上无状态的方式运行，将交互视为彼此独立，并未真正随时间增量积累知识 [1365, 1364]——尽管通过注意力标记机制实现的工作记忆进展已使灵活的记忆表示控制成为可能 [870]。这一局限阻碍了对真正终身学习的评估——终身学习是人类水平智能的基石，涉及跨多样情境与延展时间跨度的持续知识获取、保留与重用。
当试图将记忆相关性能与其他智能方面相隔离时会出现方法论问题，从而难以判定失败究竟源于记忆机制不足，还是推理能力的局限 [288]。现实应用中的动态记忆使用也对评估构成挑战，因为受控的实验室测试难以充分刻画记忆系统在复杂场景中的表现——在这些场景中，信息相关性会以不可预测的方式发生变化 [1079]。
优化策略与未来研究方向 记忆优化涵盖多种技术，旨在提升利用率的同时最小化计算开销并最大化效率。受生物学启发的遗忘机制提供了有效的优化途径，例如 MemoryBank 等框架采用艾宾浩斯遗忘曲线，依据时间因素与重要性有选择地保留或丢弃信息 [1372]。基于反思的优化则通过 Reflexion 等系统，经集成评估与自我反思进行性能评估，构建双重反馈系统，通过持续学习精炼记忆与行为 [304]。
层次化记忆结构通过多层级格式优化信息组织以实现高效检索，其例证包括具备快速记忆访问模块的基于经验的层次控制（Experience-based Hierarchical Control）框架 [868]、通过双向快–慢变量交互实现的记忆巩固过程 [63]，以及依据查询相关性动态排序记忆的自适应交叉注意力网络（Adaptive Cross-Attention Networks）[410]。
未来研究方向包括：结合参数化精度与非参数化效率的混合记忆框架 [942]、用于可扩展响应评估的自动化反馈机制 [893]、通过共享外部记忆实现协作学习的多智能体记忆系统 [306]、融合知识图谱的增强元数据学习 [896, 386]、面向特定应用的领域专用记忆架构 [507]、在非活跃期融入记忆巩固的认知启发优化 [758]，以及通过低秩适应（Low-Rank Adaptation）等技术实现参数高效的记忆更新以完成高效知识整合 [428, 256]。这些进展有望推动记忆增强的大语言模型智能体迈向更复杂、类人的认知能力，同时缓解计算与架构层面的局限；其应用可延伸至长期机器人规划、现实决策系统，以及通过流式学习场景与持续反馈集成实现的协作式 AI 助手 [1159, 1346, 1278]。
5.3. 工具集成推理（Tool-Integrated Reasoning）
工具集成推理（Tool-Integrated Reasoning）将语言模型从被动的文本生成器转变为能够动态利用工具并操控环境的主动世界交互者。该实现使模型能够通过函数调用（Function Calling）机制、集成推理框架以及复杂的环境交互能力，突破其固有局限。
5.3.1. 函数调用（Function Calling）机制
函数调用通过利用函数抽象机制的结构化输出生成，将大语言模型从生成式模型转变为交互式智能体，使其能够操控外部工具，并获取当前的、领域特定的信息以解决复杂问题 [5, 669, 335, 882, 58, 523, 1113]。
其演进始于 Toolformer 的自监督方法，该方法展示了自主 API 学习能力，并启发了 ReAct 的“思考–行动–观察”循环；随后发展为 Gorilla 等专用模型，以及 ToolLLM、RestGPT 等综合性框架，并以 OpenAI 的 JSON 标准化为标志；而 Chameleon 等进阶系统实现了多模态问答，TaskMatrix.AI 则跨领域管理各类 AI 模型 [939, 252, 654, 547, 923, 874, 875, 715, 659, 953]。
图 6：工具增强系统框架： 经函数调用机制、工具集成推理与环境交互能力，从文本生成器演进为世界交互者。
技术实现主要包括微调（主流方法，通过大规模 API 训练提供稳定能力，但资源需求较高）与提示工程（灵活、资源高效但不稳定）；诸如“逆向链”（Reverse Chain）等方法可借助提示实现 API 操作，并应对大规模工具管理中的挑战 [392, 5, 1332, 791, 144, 254]。
核心流程涵盖意图识别、函数选择、参数–值对映射、函数执行与响应生成；现代实现利用结构化的大语言模型输出与外部程序交互，而工具则包括多样接口（数字系统、草稿本、用户交互、其他大语言模型、开发者代码），需要在工具选择、参数构造与结果解析之间进行复杂导航 [1268, 669, 1141, 193, 960, 590, 910]。
训练方法与数据系统 训练方法已从基础的基于提示的方法演进为复杂的多任务学习框架；通过对专用数据集进行微调——例如 ToolLLM 与 Granite-20B-FunctionCalling 等系统——通常先使用合成的单工具数据，再辅以人工标注 [392, 5, 357, 777, 1235]。
数据生成策略包括 Weaver 基于 GPT-4 的环境合成，以及 APIGen 的分层验证流水线（格式检查、函数执行、语义验证），在数千个 API 上生成了 60,000+ 条高质量条目 [1113, 1186, 1268, 1165, 65, 1403, 749]。
工具选择增强涉及无关性感知的数据增强，包括 Hammer 的函数掩码技术、用于提升难度的 oracle 工具混合，以及用于缓解过度触发的工具意图检测合成；这些方法强调通过严格过滤与格式验证保证高质量数据 [670, 10, 357, 473, 1300, 218]。
自我改进范式通过降低对外部监督的依赖来实现，例如 JOSH 算法的稀疏奖励仿真环境，以及 TTPA 结合面向错误打分的 token 级优化；在保持通用能力的同时展现出性能提升 [579, 446, 366, 1271]。
复杂基准测试包括 API-Bank（73 个 API、314 段对话）、StableToolBench（针对 API 不稳定性的解决方案）、NesTools（嵌套工具评估）、ToolHop（995 个查询、3,912 个工具），覆盖从单工具到多跳场景 [621, 363, 377, 1264, 827, 995, 1257, 987]。
5.3.2. 工具集成推理（Tool-Integrated Reasoning）
工具集成推理（Tool-Integrated Reasoning, TIR）代表了大语言模型能力的范式性进展，通过在推理过程中与外部资源进行动态交互，应对知识过时、计算不准确以及推理浅层等根本局限 [864]。与完全依赖模型内部知识的传统推理方法不同，TIR 建立了一种协同关系：推理引导将复杂问题分解为可管理的子任务，而专用工具则确保每一步计算的准确执行 [777]。该范式超越常规的基于文本的推理，要求模型自主选择合适工具、解释中间输出，并根据实时反馈自适应地精炼其方法 [864]。
TIR 方法的演进涵盖三类主要实现，分别针对工具利用优化的不同方面。基于提示的方法通过精心设计的指令引导模型而无需额外训练，例如将数学问题分解为可执行代码、并将计算委派给 Python 解释器的方法 [155, 601]。监督微调方法通过模仿学习教会工具使用，如 ToRA 等系统聚焦数学问题求解，将自然语言推理与计算库及符号求解器相结合 [345]。强化学习方法通过结果驱动的奖励优化工具使用行为，但现有实现往往只优先最终正确性而不考虑效率，可能导致认知卸载现象——即模型过度依赖外部工具 [227]。
在操作层面，基于 TIR 的智能体充当智能编排者，系统性地将认知处理与外部资源参与交织，以实现既定目标 [1095]。该机制要求内在推理能力与外在工具利用和谐集成，以逐步完成面向目标的知识综合；智能体的执行路径可形式化为工具激活与相应信息吸收事件的结构化序列 [1095]。新兴发展已确立智能体推理（Agentic Reasoning）架构，通过引入自主部署工具的智能体放大语言模型智能，流畅编排基于网络的信息检索、计算处理以及分层的推理–记忆集成，以应对需要全面研究与级联逻辑分析的复杂挑战 [1162]。
实现框架与范式 单工具框架通过面向特定计算领域的专用实现，确立了工具集成推理的基本原则。程序辅助语言模型（Program-Aided Language Models, PAL）通过生成可执行代码并将数学计算委派给 Python 解释器，开创了问题分解策略 [309]。ToolFormer 表明，语言模型仅需极少示例即可学会外部 API 使用，并接入计算器、搜索引擎及多种工具以增强计算能力 [939]。ToRA 通过将自然语言处理与计算库及符号求解器结合推进了数学推理；ReTool 则应用强化学习优化代码解释器使用，并在自我纠错模式上展现改进 [345, 1320, 973]。Self-Edit 利用所生成代码的执行结果提升竞赛编程任务中的代码质量，采用故障感知的代码编辑器根据测试用例结果纠正错误 [1318]。
多工具协调系统应对在集成推理架构中编排异构工具的复杂性。ReAct 开创了将推理轨迹与任务特定行动交错的做法，使模型能够互补地思考与行动：推理支持计划跟踪，而行动则与外部信息源对接 [1254]。Chameleon 通过合成结合视觉模型、搜索引擎与 Python 函数的程序，并以基于大语言模型的规划器为核心，引入了即插即用的组合推理 [715]。AutoTools 建立了将原始工具文档自动转化为可执行函数的框架，降低了工具集成中的手工工程需求 [423, 960]。智能体链（Chain-of-Agents, CoA）训练模型解码带有抽象占位符的推理链，随后调用领域专用工具填补知识缺口 [600, 1337]。
基于智能体的框架代表了 TIR 系统最复杂的演进形态，超越静态提示方法，构建自主且自适应的 AI 系统。与遵循僵化模式的常规工具使用不同，智能体模型学习将思维链（Chain-of-Thought, CoT）与行动链（Chain-of-Action, CoA）模式耦合进其核心行为，从而在推理与行动之间获得更强的逻辑连贯性与更自然的过渡 [1338]。这些系统建立在基础智能体架构之上，包括将感知直接映射到行动的反应式系统、实现信念–欲望–意图（Belief-Desire-Intention, BDI）模型的慎思式系统，以及在层次结构中组合多个子系统的混合架构 [734]。
表 7：工具增强语言模型架构： 在 8 类工具上对多种方法的比较，包括搜索、计算、知识库、API、多模态、语言工具、交互环境与领域专用应用。
| 方法 | 搜索与检索 | 计算与代码执行 | 知识库与问答 | API 与外部服务 | 多模态工具 | 语言处理 | 交互环境 | 领域专用工具 |
|---|---|---|---|---|---|---|---|---|
| ReAct [1256] | ✓ | ✓ | ✓ | |||||
| Toolformer [939] | ✓ | ✓ | ✓ | ✓ | ✓ | |||
| ToolkenGPT [382] | ✓ | ✓ | ✓ | ✓ | ✓ | |||
| ToolLLM [875] | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| ToRA [345] | ✓ | |||||||
| PAL [307] | ✓ | |||||||
| HuggingGPT [953] | ✓ | ✓ | ||||||
| GPT4Tools [1234] | ✓ | |||||||
| CRITIC [344] | ✓ | ✓ | ✓ | |||||
| Chain of Code [601] | ✓ | |||||||
| TRICE [869] | ✓ | ✓ | ✓ | ✓ | ||||
| TP-LLaMA [152] | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| AlignToolLLaMA [165] | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| ReTool [274] | ✓ | |||||||
| Tool-Star [225] | ✓ | ✓ | ||||||
| ARTIST [973] | ✓ | |||||||
| Ego-R1 [1046] | ✓ | |||||||
| VTool-R1 [1164] | ✓ | |||||||
| KG-Agent [493] | ✓ | ✓ | ||||||
| CACTUS [761] | ✓ | |||||||
| MuMath-Code [1274] | ✓ | |||||||
| ToRL [627] | ✓ | |||||||
| MetaTool [458] | ✓ | ✓ | ✓ | ✓ | ||||
| ToolEyes [1262] | ✓ | ✓ | ||||||
| Graph-CoT [501] | ✓ | ✓ | ||||||
| ToolRL [864] | ✓ | ✓ | ✓ | ✓ | ||||
| LATS [1374] | ✓ | ✓ |
5.3.3. 智能体-环境交互（Agent-Environment Interaction）
强化学习方法已成为优于基于提示的方法与监督微调的工具集成替代方案，使模型能够通过探索与结果驱动的奖励自主发现最优工具使用策略 [227]。ReTool 体现了这一进展：其聚焦于面向数学推理的代码解释器优化，仅经 400 步训练即在 AIME2024 基准测试上达到 67.0% 准确率，大幅优于经充分训练后达 40.0% 准确率的基于文本的强化学习基线 [274]。这表明在决策过程中显式建模工具使用，可同时增强推理能力与训练效率。
搜索增强推理系统代表了将信息检索直接集成到推理过程中的创新：通过专用学习环境实现。Search-R1 框架训练模型在多步推理任务中动态决定何时搜索以及生成何种查询，这与传统的检索增强生成(RAG)系统不同 [984]。该架构采用专用 token 系统来组织推理与搜索过程，模型学习生成与显式搜索行动交错的推理步骤，而这些搜索行动由封装所生成查询的 token 触发 [654]。
多轮与可定制的工具调用框架应对推理过程中协调多种异构工具的复杂性。近期发展包括 VisTA 等框架，其利用强化学习使视觉智能体能够依据经验表现，从多样库中动态探索、选择并组合工具 [460]。ReVeal 通过迭代的生成–验证过程展示了自我演化的代码智能体 [512]。在多模态领域，VideoAgent 等系统将视觉–语言基础模型用作翻译与检索视觉信息的工具，并在视频理解基准测试上取得了令人印象深刻的表现 [1117, 258]。
评估与应用 对工具集成推理系统的全面评估需要专用基准测试，以衡量工具集成能力而非一般模型性能。MCP-RADAR 提供标准化评估框架，采用源自可量化性能数据的严格客观指标，并具备可扩展设计，覆盖软件工程、数学推理与一般问题求解等领域 [314]。该框架通过雷达图可视化性能，突出模型在多维度上的优势与不足，从而能够对工具集成语言模型进行系统比较，而不受实现机制限制。
面向现实世界的评估方法揭示了当前系统与人类水平能力之间的显著性能差距，为实践局限与优化机会提供了关键洞见。通用工具智能体（General Tool Agents, GTA）基准测试针对现有评估的局限：包含真实人工撰写、隐含工具使用需求的查询；部署了涵盖感知、操作、逻辑与创造力类别工具的评估平台；以及包括图像与代码片段在内的真实多模态输入 [1098]。结果表明当前大语言模型面临重大挑战，GPT-4 的完成率不到 50%。
函数调用使复杂的多智能体系统成为可能：多个大语言模型智能体通过协调的工具使用与任务分解进行协作；多智能体系统（MAS）借助并行处理、信息共享与自适应角色分配利用集体智能；而大语言模型集成则通过 DyLAN、MAD 与 MetaGPT 等框架增强了规划、专业化与任务分解能力 [243, 911, 348, 140, 631]。进阶的多智能体函数调用采用复杂编排机制，将复杂任务分解为可管理的子任务；基本方法包括将奖励机拆分为并行执行单元，每个智能体维护各自的奖励机、局部状态空间与命题；而自适应编排则能够依据上下文、响应与状态报告动态选择智能体 [39, 1056, 697, 117]。
5.4. 多智能体系统（Multi-Agent Systems）
多智能体系统（Multi-Agent Systems）代表协作智能的巅峰，使多个自主智能体能够协调与通信，以解决超出单个智能体能力的复杂问题。该实现聚焦于复杂的通信协议（Communication Protocols）、编排（Orchestration）机制以及协调（Coordination）策略，从而实现跨多样智能体架构的无缝协作。
图 7：多智能体系统框架： 面向协作式 AI 智能体系统的通信协议、编排机制与协调策略概览。
5.4.1. 通信协议（Communication Protocols）
智能体通信系统源于 20 世纪 90 年代初的知识共享计划（Knowledge Sharing Effort），通过标准化语言应对互操作性挑战，确立了自主实体协调的基本原则 [373, 93]。KQML 作为开创性的智能体通信语言（Agent Communication Language）出现，引入了将内容层、消息层与通信层相分离的多层架构，并运用言语行为理论 [373, 82, 663, 284]。FIPA ACL 则通过基于模态逻辑的语义框架、可行性前置条件与理性效应，进一步增强了这一基础 [1155, 373, 82]。
互操作性要求具备语义层面的通信能力，使跨平台智能体无需大量预通信设置即可相互理解；通过基于本体的协议形式化与语义网（Semantic Web）技术应对日益增长的异构性，并纳入针对通信脆弱性的安全机制 [486, 66, 449, 487, 792, 1063]。
当代协议生态 当代标准化协议旨在解决阻碍大语言模型智能体协作的碎片化挑战 [1244, 1137, 412]。MCP 充当“AI 的 USB-C”，通过 JSON-RPC 客户端–服务器接口标准化智能体–环境交互，使数百个服务器可覆盖多样领域，同时也引入了安全脆弱性 [934, 250, 622, 270, 15, 261, 930, 1102, 374, 1194, 301, 1016, 719, 273]。
A2A 通过基于能力的智能体卡片（Agent Cards）标准化点对点通信，经基于 JSON 的生命周期模型实现任务委派与安全协作 [622, 250, 934]。ACP 提供通用的 RESTful HTTP 通信，支持多部分消息以及同步/异步交互，并具备发现、委派与编排功能 [281, 250]。
ANP 借助 W3C 去中心化标识符与 JSON-LD 图，将互操作性扩展至开放互联网；新兴协议 AGNTCY 与 Agora 则进一步丰富了标准化生态 [250, 685, 1137]。
渐进分层策略：MCP 提供工具访问，ACP 实现消息交换，A2A 支持对等交互，ANP 扩展网络互操作性 [1015, 934]。
大语言模型增强的通信框架 大语言模型通过复杂的自然语言处理改变了智能体通信，在跨越社会科学、自然科学与工程领域的学术与工业应用中实现了前所未有的上下文敏感性 [492, 690, 504, 1099, 1179, 1136, 904, 1060, 879]。增强系统通过专用知识库、规划、记忆与内省能力展现认知协同，并支持合作式、辩论式与竞争式通信范式 [492, 360]。
通信结构涵盖分层的层次组织、去中心化的点对点网络、中心化协调以及共享消息池架构，并辅以顺序交换、通用语言接口与消息传递策略 [360, 1249, 1219, 171, 400, 491, 543, 665, 799, 949]。
框架实现支撑完整生态：AutoGen 支持动态响应生成，MetaGPT 提供共享消息池，CAMEL 提供集成编排，CrewAI 促进自适应；强化学习集成则增强奖励重设计、行动选择与策略解释 [188, 38, 119, 1004, 228, 871, 935, 958, 1273]。人–智能体通信通过灵活参与与认知多样性引入复杂交互景观，智能体可推断沟通者属性并镜像人类的交际意图 [1409, 34, 675]。
5.4.2. 编排机制（Orchestration Mechanisms）
编排机制构成多智能体系统的关键协调基础设施，管理智能体选择、上下文分发与交互流控制 [902]；通过处理用户输入、分发上下文，并基于能力评估与响应评价进行最优智能体选择，实现人类与非人类行动者之间的有效合作 [53]；同时管理消息流、确保任务推进并纠正任务偏离 [175]。进阶编排框架纳入意图识别、上下文记忆维护与任务分派组件，以跨领域专用智能体实现智能协调；其中 Swarm Agent 框架利用实时输出引导工具调用，同时应对静态工具注册表与定制通信框架的局限 [814, 267, 250]。
当代编排策略呈现不同的运行范式：先验编排通过对用户输入与智能体能力的执行前分析确定智能体选择；后验编排则将输入同时分发给多个智能体，并利用置信度指标与响应质量评估，如 3S 编排器框架所示 [901]；基于函数的编排强调从可用池中选择智能体、管理上下文信息并控制对话流 [54]；基于组件的编排采用动态规划过程，编排器依据用户指令按逻辑顺序排列组件，并将大语言模型用作组件编排工具以生成嵌入编排逻辑的工作流 [681]。
新兴编排范式包括提线木偶式编排：由中心化编排器响应不断演变的任务状态，通过基于强化学习的自适应排序与优先级动态指挥智能体；以及序列化编排：通过将协作图展开为由拓扑遍历引导的推理序列，应对协作拓扑复杂性，使编排器在每一步依据全局系统状态与任务规范选择单个智能体 [198]。
上下文管理与环境适应(Context Management and Environmental Adaptation)。 上下文是指导编排系统中智能体行动与交互的基础要素，通过全局状态维护支持运行模式多样性，同时保持应用个体性与任务执行顺序；全局状态使编排系统能够跨分布式节点跟踪任务执行进度，并为智能体提供在更广阔工作流语境中有效完成子任务所需的上下文感知 [26]。基于会话的上下文精炼界定协作范围边界，促进事件驱动编排：智能体可动态进入与退出、创建输出流并对共享会话流作出贡献；可配置会话根据用户输入或自主决策纳入智能体，从而构建能响应任务需求变化的自适应系统 [519]。
设计良好的交互结构与任务编排机制凸显了上下文在可扩展多智能体协作中的关键作用。系统根据上下文需求调整通信模式与智能体角色，通过复杂任务分解与为子任务分配合适智能体，支持面向特定任务需求的动态协作 [1137]。这种上下文适应涵盖组织与运营两个维度，使系统在保持一致性的同时，能够适应环境变化与不断演进的用户需求。
5.4.3. 协调策略(Coordination Strategies)
多智能体编排在跨复杂工作流维护事务完整性方面面临重大挑战；当代框架（包括 LangGraph、AutoGen 与 CAMEL）表现出事务支持不足：LangGraph 提供基本状态管理，但缺乏原子性保证与系统化补偿机制；AutoGen 优先支持灵活的智能体交互，却没有充分的补偿动作管理，可能在部分失败后导致系统状态不一致；验证局限亦随之显现——许多框架仅依赖大语言模型固有的自我验证能力，而未实施独立验证程序，使系统暴露于推理错误、幻觉与智能体间不一致 [128]。
上下文处理失败进一步加剧这些挑战：智能体难以维护涵盖情节性与语义性信息的长期上下文 [214, 1122]；中心编排器拓扑引入非确定性、依赖运行时的执行路径，在增强适应性的同时使异常检测更复杂，需要动态图重构而非简单路径匹配 [394]；环境错误配置与 LLM 幻觉可能干扰智能体系统，恢复不佳会导致目标偏离，并在具有分布式子任务的多智能体设置中被放大 [214, 1099]。
智能体间依赖不透明带来额外担忧：智能体可能在没有显式约束或验证层的情况下基于不一致假设或冲突数据运行，因而需要结合对编排意图与规划一致性的推理进行异常检测 [394]。应对这些挑战需要全面方案，例如提供事务支持、独立验证程序与稳健上下文保持机制的 SagaLLM 框架 [128]，以及将 Python 解释器与 LLM 智能体集成、通过多轮交互实现代码动作执行与动态修订能力的 CodeAct 等方法 [1122]。
应用与性能影响(Applications and Performance Implications)。 智能体与上下文编排在多样化应用领域展现出实用价值：医疗健康应用在专门的基于智能体架构中采用上下文切换机制，执行信息检索、问答与决策支持，利用监督智能体解释输入特征，并根据临床查询类型、用户背景与数据模态需求将子任务分配给专门智能体 [619, 760, 1059]；网络管理应用借助上下文感知编排应对复杂性挑战，为接入点(Points of Access)配备专属于特定上下文的智能体，通过包括可用服务实例与网络路径在内的上下文特定动作集，实现高效的网络动态管理 [966]。
业务流程管理与仿真通过 AgentSimulator 等平台成为重要应用领域，支持在编排与自主两种设定下进行过程行为发现与仿真：编排行为遵循全局控制流模式，活动选择依赖先前活动，智能体分配基于能力与可用性；自主行为则通过本地控制流与交接模式运行，承认协作工作中的智能体自主性 [549]。
性能影响表明，设计良好的编排通过发挥不同智能体的独特能力来提升系统有效性；研究表明，人类用户常常难以从可用集合中有效选择智能体，而自动化编排可提升整体性能 [72]，这推动了在线学习智能体能力、并在成本、能力需求与运营限制等现实约束下编排多个智能体的框架；自主性水平因实现而异，部分系统在指定阶段表现出显著自主性，在动作管理上展现与任务特异性相应的适应性，并通过上下文资源利用达到 Level 2 自主性 [466]。
6. 评估(Evaluation)
对上下文工程(Context Engineering)系统的评估提出了超越传统语言模型评估范式的前所未有的挑战。这些系统呈现复杂的多组件架构与动态、依赖上下文的行为，需要能够评估组件级诊断、基于任务的性能以及整体系统鲁棒性的综合评估框架 [841, 1141]。
上下文工程组件的异质性——涵盖检索机制、记忆系统、推理链与多智能体协调——要求评估方法既能捕捉单个组件的有效性，也能捕捉涌现的系统级行为 [314, 939]。
6.1. 评估框架与方法(Evaluation Frameworks and Methodologies)
本小节介绍用于评估上下文工程中单个组件与集成系统的综合方法。
6.1.1. 组件级评估(Component-Level Assessment)
内在评估关注孤立条件下单个组件的性能，为系统能力与失效模式提供基础洞见。
对提示工程组件而言，评估涵盖通过语义相似度指标衡量提示有效性、响应质量评估，以及跨多样化输入变体的鲁棒性测试。当前方法揭示了提示设计中的脆弱性与鲁棒性挑战，因而需要更精细的评估框架，以评估上下文校准与自适应提示优化 [1141, 669]。
长上下文处理评估需要专门指标，以衡量信息保持、位置偏差以及跨扩展序列的推理连贯性。“大海捞针(needle in a haystack)”评估范式测试模型从长上下文中检索嵌入特定信息的能力，而多文档推理任务则评估跨多个信息源的综合能力。位置插值技术与超长序列处理方法面临显著的计算挑战，限制了实际评估场景 [737, 299]。
自我上下文化机制通过元学习评估、适应速度测量以及跨多次迭代的一致性分析进行评估。包括 Self-Refine、Reflexion 与 N-CRITICS 在内的自我精炼框架表现出显著性能提升，GPT-4 通过迭代自我精炼过程约提升 20% [741, 964, 795]。多维反馈机制与基于集成的评估方法为自主演化能力提供了全面评估 [583, 710]。
结构化与关系数据集成评估考察知识图谱遍历、表格理解与数据库查询生成的准确性。然而，当前评估框架在评估结构推理能力方面存在显著局限，高质量结构化训练数据的开发仍面临持续挑战。基于 LSTM 的模型在序列信息与结构信息冲突时错误增多，凸显了需要更精细的基准测试来检验结构理解 [769, 674, 167]。
6.1.2. 系统级集成评估(System-Level Integration Assessment)
外在评估衡量下游任务的端到端性能，通过涵盖问答、推理与现实世界应用的综合基准测试，对系统效用提供整体评估。
系统级评估必须捕捉由组件交互产生的涌现行为，包括组合组件超越个体性能的协同效应，以及组件集成反而降低整体有效性的潜在干扰模式 [841, 1141]。
检索增强生成(RAG)评估通过涵盖精确率、召回率、相关性与事实准确性的综合指标，同时衡量检索质量与生成有效性。智能体化 RAG(Agentic RAG)系统引入额外复杂性，需要评估任务分解准确性、多计划选择有效性以及记忆增强规划能力。自我反思机制通过反馈环展示迭代改进，MemoryBank 实现融入 Ebbinghaus 遗忘曲线原则以增强记忆评估 [444, 166, 1372, 1192, 41]。
记忆系统评估因缺乏标准化评估框架以及当代 LLM 固有的无状态特性而面临重大困难。LongMemEval 提供 500 道精心策划的问题，评估信息提取、时间推理、跨会话推理与知识更新等基本能力。商业 AI 助手在长时间交互中表现出约 30% 的准确率下降，凸显了记忆持久性与检索有效性方面的重大不足 [1340, 1180, 463, 847, 390]。NarrativeQA、QMSum、QuALITY 与 MEMENTO 等专用基准测试应对情节记忆评估挑战 [556, 572]。
工具集成推理系统需要覆盖整个交互轨迹的综合评估，包括工具选择准确性、参数提取精度、执行成功率与错误恢复能力。MCP-RADAR 框架为软件工程与数学推理领域提供采用客观指标的标准化评估。现实世界评估揭示了显著的性能差距：在 GTA 基准测试中，GPT-4 完成任务不足 50%，而人类表现达 92% [314, 1098, 126, 939]。包括 BFCL（2,000 个测试用例）、T-Eval（553 个工具使用用例）、API-Bank（73 个 API、314 段对话）与 ToolHop（995 个查询、3,912 个工具）在内的先进基准测试，针对多轮交互与嵌套工具调用场景 [263, 363, 377, 1264, 160, 835]。
多智能体系统评估通过专门指标捕捉通信有效性、协调效率与集体结果质量，涵盖协议遵循、任务分解准确性与涌现协作行为。当代编排框架（包括 LangGraph、AutoGen 与 CAMEL）表现出事务支持不足，验证局限随之显现——系统仅依赖 LLM 自我验证能力，而无独立验证程序。上下文处理失败进一步加剧挑战，智能体难以维护涵盖情节性与语义性信息的长期上下文 [128, 394, 901]。
6.2. 基准测试数据集与评估范式(Benchmark Datasets and Evaluation Paradigms)
本小节综述专为评估上下文工程系统性能而设计的专门基准测试与评估范式。
6.2.1. 基础组件基准测试(Foundational Component Benchmarks)
长上下文处理评估采用专门基准测试套件，以检验跨扩展序列的信息保持、推理与综合。当前基准测试面临显著的计算复杂性挑战，注意力机制中的 $O(n^2)$ 缩放限制对超长序列造成实质性内存约束。位置插值与扩展技术需要精细的评估框架，能够在不同序列长度上同时评估计算效率与推理质量 [737, 299, 1236]。
包括 LongMamba 与专门位置编码方法在内的先进架构为长上下文处理展示了有前景的方向，但评估仍揭示在跨扩展序列保持连贯性方面的持续挑战。滑动注意力机制与内存高效实现的开发，需要能够同时评估计算可处理性与任务性能的综合基准测试 [1267, 351]。
结构化与关系数据集成基准测试涵盖多样化的知识表示格式与推理模式。然而，当前评估框架在评估结构推理能力方面存在局限，高质量结构化训练数据的开发仍面临持续挑战。评估必须处理序列信息与结构信息处理之间的根本张力，尤其是在这两类信息冲突的场景中 [769, 674, 167]。
6.2.2. 系统实现基准测试(System Implementation Benchmarks)
检索增强生成(RAG)评估借助应对多样化检索与生成挑战的综合基准测试套件。模块化 RAG 架构通过检索、增强与生成的专门模块展现出增强的灵活性，从而能够对单个组件及其交互进行细粒度评估。融入 GraphRAG 与 LightRAG 的图增强 RAG 系统在复杂推理场景中表现出改进性能，但评估框架必须应对图遍历与多跳推理评估的额外复杂性 [316, 973, 364]。
智能体化 RAG 系统引入精细的规划与反思机制，需要评估任务分解准确性、多计划选择有效性与迭代精炼能力。实时与流式 RAG 应用在动态信息条件下同时评估延迟与准确性方面，提出了独特的评估挑战 [444, 166, 1192]。
工具集成推理系统评估采用跨越多样化工具使用场景与复杂度层级的综合基准测试。Berkeley Function Calling Leaderboard（BFCL）提供 2,000 个测试用例，通过逐步与端到端评估衡量调用准确率、通过率与胜率，覆盖日益复杂的场景。T-Eval 贡献 553 个工具使用用例，测试多轮交互与嵌套工具调用能力 [263, 1390, 835]。包括 StableToolBench 在内的先进基准测试应对 API 不稳定挑战，NesTools 评估嵌套工具场景，ToolHop 则跨 995 个查询与 3,912 个工具评估多跳工具使用 [363, 377, 1264]。
包括 WebArena 与 Mind2Web 在内的 Web 智能体评估框架，跨 137 个网站的数千项任务提供综合评估，揭示了当前 LLM 在复杂 Web 交互能力上的显著性能差距。VideoWebArena 将评估扩展到多模态智能体，Deep Research Bench 与 DeepShop 则分别针对研究智能体与购物智能体提供专门评估 [1378, 206, 87, 482]。
多智能体系统评估采用专门框架，针对协调、通信与集体智能。然而，当前框架在跨复杂工作流的事务完整性方面面临重大挑战，许多系统缺乏针对部分失败的充分补偿机制。编排评估必须处理上下文管理、协调策略有效性，以及在变化运营条件下保持系统一致性的能力 [128, 901]。
| 发布日期 | 开源 | 方法 / 模型 | 成功率 (%) | 来源 |
|---|---|---|---|---|
| 2025-02 | × | IBM CUGA | 61.7 | [753] |
| 2025-01 | × | OpenAI Operator | 58.1 | [813] |
| 2024-08 | × | Jace.AI | 57.1 | [476] |
| 2024-12 | × | ScribeAgent + GPT-4o | 53.0 | [950] |
| 2025-01 | ✓ | AgentSymbiotic | 52.1 | [1323] |
| 2025-01 | ✓ | Learn-by-Interact | 48.0 | [998] |
| 2024-10 | ✓ | AgentOccam-Judge | 45.7 | [1231] |
| 2024-08 | × | WebPilot | 37.2 | [1331] |
| 2024-10 | ✓ | GUI-API Hybrid Agent | 35.8 | [988] |
| 2024-09 | ✓ | Agent Workflow Memory | 35.5 | [1144] |
| 2024-04 | ✓ | SteP | 33.5 | [979] |
| 2025-06 | ✓ | TTI | 26.1 | [951] |
| 2024-04 | ✓ | BrowserGym + GPT-4 | 23.5 | [238] |
表 8：WebArena [1378] 排行榜：表现最佳模型及其成功率与可用性状态。（原文表见 PDF 第 48 页）
6.3. 评估挑战与新兴范式(Evaluation Challenges and Emerging Paradigms)
本小节识别评估方法中的当前局限，并探讨更有效评估的新兴方法。
6.3.1. 方法局限与偏差(Methodological Limitations and Biases)
传统评估指标从根本上不足以捕捉上下文工程系统所展现的细微、动态行为。BLEU、ROUGE 与困惑度等静态指标最初为更简单的文本生成任务设计，无法评估复杂推理链、多步交互与涌现系统行为。多组件系统固有的复杂性与相互依赖造成归因挑战，使隔离故障与识别根因在计算与方法上变得难以处理。未来指标必须演进，不仅捕捉任务成功，还要捕捉底层推理过程的质量与鲁棒性，尤其是在需要组合泛化与创造性问题解决的场景中 [841, 1141]。
记忆系统评估因缺乏标准化基准测试以及当前 LLM 的无状态性质而面临特殊挑战。自动化记忆测试框架必须解决隔离问题——不同记忆测试阶段无法有效分离，导致评估结果不可靠。商业 AI 助手在持续交互中表现出显著性能退化，准确率下降可达 30%，凸显了当前评估方法的关键缺口，并指向需要能够随时间跟踪记忆保真度的纵向评估框架 [1340, 1180, 463]。
工具集成推理系统评估揭示了当前系统与人类水平能力之间的显著性能差距。GAIA 基准测试表明，人类在通用助手任务上达到 92% 准确率，而 GPT-4 等先进模型仅达 15%，表明当前评估框架与系统能力存在根本局限 [778, 1098, 126]。评估框架必须应对多工具协调、错误恢复以及跨多样化运营上下文的自适应工具选择的复杂性 [314, 939]。
6.3.2. 新兴评估范式(Emerging Evaluation Paradigms)
自我精炼评估范式利用迭代改进机制，跨多个精炼周期评估系统能力。包括 Self-Refine、Reflexion 与 N-CRITICS 在内的框架通过多维反馈与基于集成的评估方法展现出显著性能提升。GPT-4 通过自我精炼过程约提升 20%，凸显了跨多次迭代周期而非单次评估系统的重要性。然而，未来的关键挑战在于评估元学习能力本身——不仅看系统是否改进，还要看它随时间学习精炼策略的效率与鲁棒性 [741, 964, 795, 583]。
多方面反馈评估纳入正确性、相关性、清晰度与鲁棒性等多样化反馈维度，为系统输出提供综合评估。自我奖励机制使自主演化与元学习评估成为可能，允许系统通过迭代精炼发展日益精细的评估标准 [710]。
批评引导评估采用专门的批评模型对系统输出提供详细反馈，从而对推理质量、事实准确性与逻辑一致性进行细粒度评估。这些方法通过提供可适应多样化任务需求与输出格式的上下文感知、内容感知评估，弥补了传统指标的局限 [795, 583]。
编排评估框架通过纳入事务完整性评估、上下文管理评估与协调策略有效性测量，应对多智能体协调的独特挑战。包括 SagaLLM 在内的先进框架提供事务支持与独立验证程序，以弥补仅依赖 LLM 自我验证能力的系统局限 [128, 394]。
6.3.3. 安全性与鲁棒性评估(Safety and Robustness Assessment)
面向安全的评估纳入综合鲁棒性测试、对抗攻击抵抗力与对齐评估，以确保上下文工程系统的负责任开发。必须特别关注可在长时间内自主运行的智能体系统评估，因为这些系统提出了传统评估框架无法充分应对的独特安全挑战 [973, 364]。
鲁棒性评估必须通过综合压力测试协议，评估系统在分布偏移、输入扰动与对抗条件下的性能。多智能体系统在协调失败场景中面临额外挑战，部分系统故障可级联至整个智能体网络。评估框架必须处理优雅降级策略、错误恢复协议，以及在不利条件下保持系统功能的能力。除预定义失效模式外，未来评估还必须应对对“未知的未知”——高度复杂、自主多智能体系统中涌现且不可预测的故障级联——的韧性评估 [128, 394]。
对齐评估通过专门评估框架衡量系统对预期行为的遵循、价值一致性与有益结果优化。上下文工程系统因其动态适应能力与跨多组件的复杂交互模式，提出了独特的对齐挑战。长期评估必须检验系统在经延长运营周期适应与演化时，是否保持有益行为 [901]。
展望未来，上下文工程系统的评估需要从静态基准测试转向动态、整体性评估的范式转变。未来框架必须超越衡量任务成功，进而评估面向新问题的组合泛化，并在交互环境中跟踪长期自主性。与 AI 能力共同演进的“活”基准测试的发展，以及社会技术与经济指标的整合，对于确保这些先进系统在现实世界应用中不仅强大，而且可靠、高效并与人类价值观对齐至关重要 [314, 1378, 1340]。
上下文工程系统的评估格局随新架构、能力与应用的出现而持续快速演进。未来评估范式必须在应对日益增长的系统复杂性的同时，为系统改进与部署决策提供可靠、全面且可操作的洞见。从组件级评估到系统范围鲁棒性测试等多种评估方法的整合，是确保上下文工程系统在现实世界应用中可靠部署的关键研究优先事项 [841, 1141]。
7. 未来方向与开放挑战(Future Directions and Open Challenges)
上下文工程正处于关键拐点：基础进展与新兴应用需求交汇，在创造前所未有创新机遇的同时，也揭示了需要跨多个维度持续研究努力的根本挑战 [841, 1141]。
随着该领域从孤立组件开发转向集成系统架构，研究挑战的复杂性呈指数增长，要求能够连接理论计算机科学、实用系统工程与领域专业知识的跨学科方法 [314, 939]。
本节系统考察将界定未来十年上下文工程演进的关键研究方向与开放挑战(Open Challenges)。
7.1. 基础研究挑战(Foundational Research Challenges)
本小节考察为使上下文工程系统超越当前局限而必须应对的核心理论与计算挑战。
7.1.1. 理论基础与统一框架(Theoretical Foundations and Unified Frameworks)
上下文工程目前缺乏连接不同技术并提供原则性设计指南的统一理论基础(Theoretical Foundations)，这是限制系统化进展与最优系统开发的关键研究缺口。
缺乏刻画不同架构配置下上下文工程能力、局限与最优设计原则的数学框架，阻碍了基础理解与实际优化 [1141, 669, 841, 314]。
对上下文工程系统的信息论分析，需要全面研究最优上下文分配策略、信息冗余量化，以及上下文窗口内的根本压缩极限。当前方法缺乏确定最优上下文构成的原则性手段，导致资源利用欠优与性能下降。研究必须建立上下文效率的数学界限，开发上下文选择的优化算法，并创建用于预测不同上下文配置下系统行为的理论框架 [737, 299]。
对上下文工程系统的组合性理解，需要描述单个组件在集成架构中如何交互、干扰与协同的形式模型。由组件交互产生的复杂行为涌现，要求通过实证研究与理论建模方法进行系统调查。多智能体编排在开发用于预测协调有效性与涌现协作行为的数学框架方面，提出了特别挑战 [128, 901]。
7.1.2. 缩放定律与计算效率(Scaling Laws and Computational Efficiency)
LLM 卓越理解能力与其显著生成局限之间的根本不对称，是上下文工程研究中最关键的挑战之一。
这种理解–生成差距体现在多个维度，包括长篇输出连贯性、事实一致性维护与规划精细度，需要研究这些局限究竟源于架构约束、训练方法，还是根本的计算边界 [841, 1141]。
长篇生成能力需要系统研究能够在数千个 token 上保持连贯性、同时维护事实准确性与逻辑一致性的规划机制。当前系统在扩展生成任务中表现出显著性能退化，凸显了需要超越传统 Transformer 范式的架构创新。包括 Mamba 在内的状态空间模型通过线性缩放特性展现出更高效长序列处理的潜力，但当前实现仍需大量发展，才能在多样化任务上匹配 Transformer 性能 [737, 1267, 351, 220]。
上下文扩展效率面临根本性的计算挑战：当前注意力机制随序列长度呈二次方（$O(n^{2})$）扩展，对超长序列造成难以承受的内存与计算开销。滑动注意力机制与内存高效实现是有前景的方向，但仍需大量研究以同时解决计算可行性与推理质量保持问题 [299, 1236, 351]。位置插值与扩展技术也需进一步发展，以便在处理超出当前架构限制的序列时仍能保持位置理解与连贯性。
7.1.3. 多模态集成（Multi-Modal Integration）与表示
在上下文工程系统中融合多样模态，在表示学习、跨模态推理与统一架构设计方面提出了根本性挑战。当前方法通常采用模态专用编码器，跨模态交互有限，难以捕捉复杂多模态理解所特有的丰富相互依赖关系。VideoWebArena 展示了多模态智能体评估的复杂性，揭示了当前系统在同时处理视频、音频与文本时存在显著性能差距 [482]。
除这些感知模态之外，上下文工程还必须处理更抽象的信息形式，例如图（graph）——其结构语义无法被语言模型直接解释。捕捉图结构中编码的高层含义带来独特挑战，包括将图表示与语言模型嵌入对齐，以及高效表达图拓扑。近期工作如 GraphGPT [1032] 与 GraphRAG [248] 试图通过跨模态对齐策略弥合这一差距，另一些研究则探索将图转化为自然语言描述以促进模型理解 [266, 323]。Bi 等人 [75] 进一步提出分而治之的方法来编码文本属性异构网络，以应对上下文长度限制并实现有效的链接预测。因此，图推理成为上下文工程中的核心难点，要求模型在原始模态之外驾驭复杂关系结构。
跨多模态上下文的时序推理需要能够在扩展序列中追踪对象持续性、因果关系与时序动态的复杂架构。包括 WebArena 在内的 Web 智能体框架展示了在涉及多样模态与动态内容的复杂多步交互中保持连贯理解的挑战。当前系统在协调多模态信息处理与动作规划、执行方面表现出显著局限 [1378, 206]。
跨模态对齐与一致性是持续存在的挑战，需确保从不同模态提取的信息在事实上保持一致、在语义上保持连贯。Deep Research Bench 评估表明，当前多模态智能体在需要综合文本、视觉与结构化数据源的复杂研究任务上仍有困难，凸显了对更精细对齐机制的需求 [87]。
7.2. 技术创新机遇
本小节探讨有望增强上下文工程能力的新兴技术路径与架构创新。
7.2.1. 下一代架构（Next-Generation Architectures）
超越传统 Transformer 范式的架构创新，为解决上下文工程系统当前局限提供了有前景的方向。包括 LongMamba 在内的状态空间模型通过线性扩展特性与改进的内存利用，展现了更高效长序列处理的潜力，但当前实现仍需大量发展才能在多样任务上匹敌 Transformer 性能 [1267, 737]。专用位置编码方法与参数高效架构为在保持计算可行性的同时扩展到超长序列提供了机遇 [351, 299]。
记忆增强架构需超越当前外部记忆机制，以实现更精细的长期记忆组织、层次化记忆结构与自适应记忆管理策略。融入艾宾浩斯遗忘曲线（Ebbinghaus Forgetting Curve）原理的 MemoryBank 实现展示了记忆持久性方面的有前景方法，但仍需大量研究以应对当前大语言模型（LLM）根本性的无状态特性 [1372, 1340, 1180, 819, 1211]。开发能够在扩展交互中保持连贯长期上下文的情景记忆系统，是一项关键架构挑战 [463, 847, 397]。
模块化与组合式架构通过专用组件集成实现灵活的系统构建，同时保持整体系统连贯性。模块化检索增强生成(RAG)架构通过检索、增强与生成的专用模块展现更强灵活性，从而可对各个组件进行细粒度优化。包括 GraphRAG 与 LightRAG 在内的图增强方法展示了将结构化知识表示与神经处理相结合的潜力 [316, 973, 364]。
7.2.2. 高级推理与规划（Advanced Reasoning and Planning）
上下文工程系统需要增强的推理能力，涵盖因果推理、反事实思维、时序推理以及跨扩展上下文的类比推理。当前系统在需要整合多证据源、考虑替代情景并在复杂推理链中保持逻辑一致性的精细推理模式上能力有限 [1141, 841]。
多步规划与执行能力是关键推进方向，使系统能够分解复杂任务、制定执行策略、监控进度，并根据中间结果调整计划。智能体式 RAG（Agentic RAG）系统展示了复杂的规划与反思机制，需要整合任务分解、多计划选择与迭代精炼能力。然而，当前实现在扩展规划时域上保持连贯性以及适应动态信息条件方面仍面临重大挑战 [444, 166, 1192]。
工具集成推理代表一种范式性进展，要求在推理过程中与外部资源进行动态交互。GAIA 基准测试显示了巨大的性能差距：人类达到 92% 准确率，而先进模型仅约 15%，凸显了当前推理与规划能力的根本局限 [778, 1098, 126]。先进的工具集成必须解决自主工具选择、参数提取、多工具协调以及在多样操作情境下的错误恢复等问题 [314, 939]。
7.2.3. 复杂上下文组织与图问题求解
图推理是上下文工程中的根本挑战，要求系统在互联元素间驾驭复杂结构关系的同时保持语义理解。图—语言模型集成的近期进展展示了多种范式：纳入图专用组件的专用架构方法，以及将图结构转化为自然语言表示的基于文本的编码策略 [1093, 1031]。
架构集成方法包括 GraphGPT，它采用双阶段指令微调，通过自监督图匹配将图结构信息与语言 token 对齐 [1031, 747]。该框架引入经 Graph Instruction Tuning 精炼的专用 GraphTokens，并利用轻量级图—文本对齐投影器在文本处理与结构处理模态之间转换 [1279, 278]。在指令微调范式基础上，GraphWiz 进一步引入 DPO 以增强推理可靠性，在多样图任务上达到 65% 的平均准确率，显著优于 GPT-4 的 43.8% [145]。思维链蒸馏机制可增强逐步推理性能 [1147, 1401]。强化学习（RL）是另一有前景的方向，如 G1 所示：该工作在 Erdős 数据集上对 LLM 进行图论合成任务训练（涵盖 50 类多样任务），以 3B 参数模型取得强零样本泛化，并优于显著更大规模的模型 [361]。
基于文本的编码方法通过少样本提示与思维链推理，在无需修改架构的情况下将图结构转化为自然语言描述 [266, 196]。这些方法引入多样图描述模板，通过多种语义解释对结构元素进行情境化 [944, 722]。近期工作研究了图描述顺序对 LLM 性能的影响，发现序列呈现方式显著影响模型理解与推理准确率 [323]。基准评估亦在扩展：GraphArena 同时提供多项式时间任务与 NP 完全挑战，并采用严格评估框架，将输出分类为正确、次优、幻觉或缺失 [1033]。结合 NLGraph 与 GraphDO 等现有基准，这些评估揭示了简单连通性问题与最大流计算等复杂任务之间的显著性能差距 [1093, 903, 323]。
当前实现在扩展到大规模结构、跨多跳关系保持一致性以及对新拓扑泛化等方面仍面临挑战：基于文本的方法以降低结构精度为代价提供可解释性，而专用架构通过增加复杂度获得更强性能 [897, 1109]。包括 InstructGraph 与 GraphAdapter 在内的新兴混合方法试图通过结构化格式言语化器与基于 GNN 的适配器弥合这些范式，但在处理动态结构与关系的时序演化方面局限依然存在 [265]。展望未来，以联想网络而非碎片化搜索来组织信息、从中心节点向外扩散以发现实体间潜在联系的广泛连接范式，或可代表面向复杂上下文组织的下一代 RAG 系统 [131]。
7.2.4. 智能上下文组装与优化
能够从可用组件中智能组装上下文的自动化上下文工程系统，是关键研究前沿，需要发展上下文优化算法、自适应选择策略与可学习的组装函数。当前方法严重依赖启发式手段与领域专用工程，限制了跨多样应用的可扩展性与最优性 [1141, 669]。
自精炼机制通过迭代改进过程展现出智能上下文优化的巨大潜力。Self-Refine、Reflexion 与 N-CRITICS 框架取得了显著性能提升，其中 GPT-4 经迭代精炼约提升 20%。然而，这些方法仍需在跨多样上下文的自主演化与元学习优化策略方面进一步推进 [741, 964, 795, 583]。
纳入正确性、相关性、清晰度与鲁棒性等多样反馈维度的多维反馈机制，为上下文优化提供了有前景的方向。自我奖励机制使自主演化成为可能，但研究仍须回答关于最优适应速率、稳定性—可塑性权衡，以及在变化运行条件下保留有益适应等根本问题 [710]。
7.3. 应用驱动的研究方向
本小节讨论由真实部署需求与领域专用应用所引发的研究挑战。
7.3.1. 领域特化与适应
上下文工程系统需要面向医疗健康、法律分析、科学研究、教育与工程应用等多样领域的精细特化机制；各领域在知识整合、推理模式、安全考量与法规遵从方面具有独特要求。领域专用优化需要研究迁移学习策略、领域适应技术以及在增强领域性能的同时保持通用能力的专用训练范式 [1141, 669]。
科学研究应用要求对复杂技术内容、数学表达式、实验数据与理论框架进行精细推理，同时维持严格的准确性标准。Deep Research Bench 评估揭示了当前系统在开展需要跨多信息源综合并对技术内容进行推理的复杂研究任务方面存在显著挑战。研究须解决符号推理与神经方法的融合，以及领域专用知识库的纳入 [87]。
医疗健康应用要求全面的安全评估框架、法规遵从机制、隐私保护协议，以及与现有临床工作流的集成，同时满足可解释性与可审计性要求。医学上下文工程必须应对敏感信息处理、确保临床准确性、支持诊断推理，以及在复杂医疗生态中维护患者隐私等挑战。当前评估框架揭示了医学推理能力与安全评估方法学方面的显著缺口 [390]。
7.3.2. 大规模多智能体协调
将多智能体上下文工程系统扩展到数百乃至数千参与智能体，需要发展分布式协调机制、高效通信协议与层次化管理结构，以在保持系统连贯性的同时赋予局部自主性。研究必须应对大规模智能体群体中的分布式共识、容错与涌现行为预测等根本挑战 [243, 140]。
通信协议标准化是关键研究前沿；新兴协议包括 MCP（“AI 的 USB-C”）、A2A（Agent-to-Agent）、ACP（Agent Communication Protocol）与 ANP（Agent Network Protocol），表明需要统一框架以实现跨多样智能体生态的互操作性。然而，当前实现面临安全漏洞与可扩展性局限，大规模部署前必须加以解决 [37, 1015, 468, 1, 250, 934, 622]。
编排挑战——包括事务完整性、上下文管理与协调策略有效性——是大规模多智能体部署的重大障碍。包括 LangGraph、AutoGen 与 CAMEL 在内的当代框架表现出事务支持不足与验证局限，致使系统只能依赖 LLM 的自我验证能力。先进协调框架必须应对部分失败的补偿机制，并在变化运行条件下保持系统连贯性 [128, 394, 901]。
7.3.3. 人机协作（Human-AI Collaboration）与集成
精细的人机协作框架需要深入理解人类认知过程、沟通偏好、信任动态与协作模式，以构建发挥互补优势的有效混合团队。研究必须探究最优任务分配策略、通信协议，以及人与 AI 系统之间共享心智模型的构建 [1141, 841]。
Web 智能体评估框架揭示了人机协作中的显著挑战，尤其是在需要持续交互与协调的复杂任务情景中。WebArena 与 Mind2Web 表明，当前系统在跨多样网站的多步交互上仍有困难，凸显了协作任务执行方面的根本缺口。先进接口需要研究情境感知适应与个性化机制，以提升人机团队绩效 [1378, 206]。
信任校准与透明机制是确保人类对 AI 系统形成恰当依赖、同时保持人类能动性与决策权威的关键研究领域。评估框架必须涵盖解释生成、不确定性传达与置信度校准，以支持协作情景中的知情人类决策。GAIA 等基准所揭示的巨大性能差距，凸显了开发能够有效传达自身局限与能力之系统的重要性 [778, 1098]。
7.4. 部署与社会影响考量
本小节审视在规模化部署上下文工程系统时，为确保负责任且有益的结果所需的关键考量。
7.4.1. 可扩展性与生产部署
上下文工程系统的生产部署需要在多个维度应对可扩展性挑战，包括计算资源管理、延迟优化、吞吐量最大化与成本效率，同时在多样运行条件下保持一致性能。当前注意力机制的 $O(n^{2})$ 扩展局限对在生产环境中部署超长上下文系统构成重大障碍，因而需要推进内存高效架构与滑动注意力机制 [299, 1236]。
随着上下文工程系统在跨领域决策过程中承担日益重要的角色，可靠性与容错机制变得至关重要。多智能体编排框架在跨复杂工作流保持事务完整性方面面临特殊挑战，当前系统缺乏针对部分失败的充分补偿机制。研究必须应对优雅降级策略、错误恢复协议以及在不利条件下维持系统功能的冗余机制 [128, 394]。
可维护性与演化挑战需要研究系统版本控制、向后兼容、持续集成协议以及支持在不中断已部署服务的前提下持续改进的自动化测试框架。由于当前 LLM 的无状态特性，以及缺乏针对长期记忆持久性与检索效率的标准化基准，记忆系统实现面临额外挑战 [1340, 1180]。
7.4.2. 安全性、安保与鲁棒性（Safety, Security, and Robustness）
全面的安全评估需要发展能够识别上下文工程系统全能力谱系中潜在失效模式、安全违规与非预期行为的评估框架。智能体系统因其自主运行能力以及在扩展运行周期中的复杂交互模式而带来特殊安全挑战 [973, 364]。
安保考量涵盖对对抗攻击、数据投毒、提示注入、模型窃取与隐私侵犯的防护，同时保持系统功能与可用性。包括 MCP、A2A 与 ACP 在内的多智能体通信协议引入了必须在保持互操作性与功能的同时加以解决的安全漏洞。研究必须发展防御机制与检测系统，以应对分布式智能体网络中不断演变的威胁态势 [250, 934]。
对齐（alignment）与价值规约挑战需要研究如何确保上下文工程系统按预期目标行事，同时避免规约博弈、奖励黑客与目标错位。由于其动态适应能力以及跨多组件的复杂交互模式，上下文工程系统呈现独特的对齐挑战。评估框架所揭示的巨大性能差距，凸显了发展能够在系统演化与适应过程中保持有益行为的鲁棒对齐机制的重要性 [778, 128]。
7.4.3. 伦理考量（Ethical Considerations）与负责任开发
偏见缓解与公平性评估需要全面评估框架，以识别并应对跨不同人口群体、应用领域与用例的系统性偏见，同时保持系统性能与效用。研究必须探究训练数据、模型架构与部署情境中的偏见来源，并发展针对根源而非表象的缓解策略 [1141, 841]。
隐私保护机制必须应对敏感信息处理、防止数据泄露以及在赋能有益系统能力的同时维护用户隐私等挑战。记忆系统因其持久信息存储与检索能力而面临特殊隐私挑战，需要安全记忆管理与选择性遗忘机制的先进框架 [1340, 463]。
透明性与问责框架需要发展解释系统、审计机制与治理结构，以便对上下文工程系统进行负责任的监督，同时支持创新与有益应用。GAIA 等评估框架所揭示的巨大性能差距（人类 92% 对 AI 15%）凸显了透明传达能力与为已部署系统设定恰当期望的重要性 [778, 1098]。
上下文工程（Context Engineering）的未来将取决于我们能否通过持续、协作的研究努力应对这些相互关联的挑战，并将技术创新与社会考量相衔接。成功需要持续投入基础研究、跨学科协作以及负责任的开发实践，以确保上下文工程系统在日益融入关键社会功能时保持有益、可靠并与人类价值观对齐 [841, 1141, 314]。
8. 结论（Conclusion）
本综述首次对上下文工程作为一门形式化学科进行了全面审视：该学科系统性地设计、优化并管理面向 LLM 的信息载荷。通过对逾 1400 篇研究论文的分析，我们将上下文工程确立为开发复杂 AI 系统的关键基础——此类系统能够有效整合外部知识、维持持久记忆，并与复杂环境动态交互。
我们的主要贡献在于提出统一的分类框架，将上下文工程技术组织为基础组件（Foundational Components：上下文检索与生成、上下文处理、上下文管理）与系统实现（System Implementations：检索增强生成、记忆系统、工具集成推理、多智能体系统）。该框架展示了核心技术能力如何整合为应对真实世界需求的复杂架构。
通过这一系统性审视，我们识别出若干关键洞见。第一，我们观察到 LLM 在理解复杂上下文方面的卓越能力与其生成同等精细输出方面的局限之间存在根本不对称。这一理解—生成鸿沟是该领域面临的最关键挑战之一。第二，我们的分析揭示了日益精细的集成模式：多种技术协同组合，创造出超越各组件之和的能力。第三，我们观察到向模块化与组合性发展的清晰趋势，使架构能够灵活适应多样应用，同时保持系统连贯性。我们所识别的评估挑战凸显了对能够刻画上下文工程系统所表现的复杂、动态行为的全面评估框架的需求。传统评估方法对整合多组件、展现自适应行为并在扩展时域上运行的系统已显不足。对未来研究方向的审视揭示了显著机遇，包括开发高效处理长上下文的下一代架构、创建智能上下文组装系统，以及推进多智能体协调机制。关键挑战横跨理论基础、技术实现与实际部署，包括缺乏统一理论框架、扩展局限以及安全考量。
展望未来，随着领域迈向复杂的多组件系统，上下文工程有望在 AI 发展中扮演日益核心的角色。上下文工程的跨学科本质要求跨越计算机科学、认知科学、语言学与领域专业知识的协作研究路径。
随着 LLM 持续演进，上下文工程的根本洞见——即 AI 系统性能根本上由上下文信息所决定——将继续居于人工智能发展的中心。本综述既提供当前状态的全面快照，也为未来研究提供路线图，将上下文工程确立为一门具有自身原则、方法与挑战的独立学科，以促进创新并支持情境感知 AI 系统的负责任发展。
致谢（Acknowledgments）
本综述代表一项持续努力，旨在全面描绘面向大语言模型的上下文工程这一快速演进的图景。鉴于该领域的动态特性以及新进展不断涌现，我们承认：尽管已尽最大努力，仍可能无意中遗漏或未充分呈现部分近期工作或新兴趋势。我们欢迎研究社区反馈，以改进本工作的未来迭代。我们感谢更广泛研究社区的奠基性贡献，使本综述成为可能。若无研究社区与开源社区的宝贵支持，本工作将无法完成——双方在开发框架、工具与资源方面的协作努力显著推进了上下文工程领域。我们特别感谢 Long Chain-of-Thought [149] 与 AI4Research [151] 项目团队出色的模板设计与可视化，它们显著提升了本综述的呈现质量。我们对他们为研究社区所作的周到贡献深表感谢。
参考文献
Acknowledgments
This survey represents an ongoing effort to comprehensively map the rapidly evolving landscape of Context
Engineering for Large Language Models. Given the dynamic nature of this field, with new developments
emerging continuously, we acknowledge that despite our best efforts, some recent works or emerging trends
may have been inadvertently overlooked or underrepresented. We welcome feedback from the research
community to help improve future iterations of this work. We are grateful to the broader research community
whose foundational contributions have made this survey possible. This work would not have been achievable
without the invaluable support of both the research community and the open-source community, whose
collaborative efforts in developing frameworks, tools, and resources have significantly advanced the field of
Context Engineering. We extend special gratitude to the teams behind the Long Chain-of-Thought [149] and
AI4Research [151] projects for their excellent template designs and visualizations, which have significantly
enhanced the presentation quality of this survey. Their thoughtful contributions to the research community
are deeply appreciated.
References
[1] Anp-agent communication meta-protocol specification(draft). https://
agent-network-protocol.com/specs/communication.html. [Online; accessed 17-
July-2025].
[2] S. A. Automating human evaluation of dialogue systems. North American Chapter of the Association
for Computational Linguistics, 2022.
[3] Samir Abdaljalil, Hasan Kurban, Khalid A. Qaraqe, and E. Serpedin. Theorem-of-thought: A multi-
agent framework for abductive, deductive, and inductive reasoning in language models. arXiv
preprint, 2025.
[4] Abdelrahman Abdallah, Bhawna Piryani, Jamshid Mozafari, Mohammed Ali, and Adam Jatowt.
Rankify: A comprehensive python toolkit for retrieval, re-ranking, and retrieval-augmented genera-
tion, arXiv preprint arXiv:2502.02464, 2025. URL https://arxiv.org/abs/2502.02464v3.
[5] Ibrahim Abdelaziz, Kinjal Basu, Mayank Agarwal, Sadhana Kumaravel, Matt Stallone, Rameswar
Panda, Yara Rizk, G. Bhargav, M. Crouse, Chulaka Gunasekara, S. Ikbal, Sachin Joshi, Hima P.
Karanam, Vineet Kumar, Asim Munawar, S. Neelam, Dinesh Raghu, Udit Sharma, Adriana Meza
Soria, Dheeraj Sreedhar, P. Venkateswaran, Merve Unuvar, David Cox, S. Roukos, Luis A. Lastras, and
P. Kapanipathi. Granite-function calling model: Introducing function calling abilities via multi-task
learning of granular tasks. Conference on Empirical Methods in Natural Language Processing, 2024.
[6] D. Acharya, Karthigeyan Kuppan, and Divya Bhaskaracharya. Agentic ai: Autonomous intelligence
for complex goals—a comprehensive survey. IEEE Access, 2025.
[7] Manoj Acharya, Kushal Kafle, and Christopher Kanan. Tallyqa: Answering complex counting
questions. AAAI Conference on Artificial Intelligence, 2018.
[8] Shantanu Acharya, Fei Jia, and Boris Ginsburg. Star attention: Efficient llm inference over
long sequences, arXiv preprint arXiv:2411.17116, 2024. URL https://arxiv.org/abs/2411.
17116v3.
[9] Emre Can Acikgoz, Jeremy Greer, Akul Datta, Ze Yang, William Zeng, Oussama Elachqar, Emmanouil
Koukoumidis, Dilek Hakkani-Tur, and Gokhan Tur. Can a single model master both multi-turn
conversations and tool use? coalm: A unified conversational agentic language model, arXiv preprint
arXiv:2502.08820, 2025. URL https://arxiv.org/abs/2502.08820v3.
[10] Emre Can Acikgoz, Cheng Qian, Hongru Wang, Vardhan Dongre, Xiusi Chen, Heng Ji, Dilek Hakkani-
Tur, and Gokhan Tur. A desideratum for conversational agents: Capabilities, challenges, and
future directions, arXiv preprint arXiv:2504.16939, 2025. URL https://arxiv.org/abs/2504.
16939v1.
[11] Anum Afzal, Juraj Vladika, Gentrit Fazlija, Andrei Staradubets, and Florian Matthes. Towards opti-
mizing a retrieval augmented generation using large language model on academic data. International
Conference on Natural Language Processing and Information Retrieval, 2024.
[12] Ankush Agarwal, Sakharam Gawade, A. Azad, and P. Bhattacharyya. Kitlm: Domain-specific
knowledge integration into language models for question answering. ICON, 2023.
[13] Oshin Agarwal, Heming Ge, Siamak Shakeri, and Rami Al-Rfou. Large scale knowledge graph based
synthetic corpus generation for knowledge-enhanced language model pre-training. arXiv preprint,
2020.
[14] Monica Agrawal, S. Hegselmann, Hunter Lang, Yoon Kim, and D. Sontag. Large language models
are few-shot clinical information extractors. Conference on Empirical Methods in Natural Language
Processing, 2022.
[15] Arash Ahmadi, S. Sharif, and Yaser Mohammadi Banadaki. Mcp bridge: A lightweight, llm-agnostic
restful proxy for model context protocol servers, arXiv preprint arXiv:2504.08999, 2025. URL
https://arxiv.org/abs/2504.08999v1.
[16] J. Ainslie, J. Lee-Thorp, Michiel de Jong, Yury Zemlyanskiy, Federico Lebr’on, and Sumit K. Sanghai.
Gqa: Training generalized multi-query transformer models from multi-head checkpoints. Conference
on Empirical Methods in Natural Language Processing, 2023.
[17] Adel Al-Jumaily. Multi-agent system concepts theory and application phases. arXiv preprint, 2006.
[18] Faisal Al-Khateeb, Nolan Dey, Daria Soboleva, and Joel Hestness. Position interpolation improves
alibi extrapolation. arXiv preprint, 2023.
[19] Jean-Baptiste Alayrac, Jeff Donahue, Pauline Luc, Antoine Miech, Iain Barr, Yana Hasson, Karel Lenc,
A. Mensch, Katie Millican, Malcolm Reynolds, Roman Ring, Eliza Rutherford, Serkan Cabi, Tengda
Han, Zhitao Gong, Sina Samangooei, Marianne Monteiro, Jacob Menick, Sebastian Borgeaud, Andy
Brock, Aida Nematzadeh, Sahand Sharifzadeh, Mikolaj Binkowski, Ricardo Barreira, O. Vinyals,
Andrew Zisserman, and K. Simonyan. Flamingo: a visual language model for few-shot learning.
Neural Information Processing Systems, 2022.
[20] Stefano V. Albrecht and P. Stone. Autonomous agents modelling other agents: A comprehensive
survey and open problems. Artificial Intelligence, 2017.
[21] Buthayna AlMulla, Maram Assi, and Safwat Hassan. Understanding the challenges and promises of
developing generative ai apps: An empirical study, arXiv preprint arXiv:2506.16453, 2025. URL
https://arxiv.org/abs/2506.16453v2.
[22] Reem S. Alsuhaibani, Christian D. Newman, M. J. Decker, Michael L. Collard, and Jonathan I. Maletic.
On the naming of methods: A survey of professional developers. International Conference on Software
Engineering, 2021.
[23] Francesco Alzetta, P. Giorgini, A. Najjar, M. Schumacher, and Davide Calvaresi. In-time explainability
in multi-agent systems: Challenges, opportunities, and roadmap. EXTRAAMAS@AAMAS, 2020.
[24] Kenza Amara, Lukas Klein, Carsten T. Lüth, Paul F. Jäger, Hendrik Strobelt, and Mennatallah El-
Assady. Why context matters in vqa and reasoning: Semantic interventions for vlm input modalities,
arXiv preprint arXiv:2410.01690v1, 2024. URL https://arxiv.org/abs/2410.01690v1.
[25] Xavier Amatriain. Prompt design and engineering: Introduction and advanced methods, arXiv
preprint arXiv:2401.14423, 2024. URL https://arxiv.org/abs/2401.14423v4.
[26] Zahra Aminiranjbar, Jianan Tang, Qiudan Wang, Shubha Pant, and Mahesh Viswanathan. Dawn:
Designing distributed agents in a worldwide network, arXiv preprint arXiv:2410.22339, 2024. URL
https://arxiv.org/abs/2410.22339v3.
[27] Chenxin An, Jun Zhang, Ming Zhong, Lei Li, Shansan Gong, Yao Luo, Jingjing Xu, and Lingpeng
Kong. Why does the effective context length of llms fall short? International Conference on Learning
Representations, 2024.
[28] Kaikai An, Fangkai Yang, Liqun Li, Junting Lu, Sitao Cheng, Shuzheng Si, Lu Wang, Pu Zhao, Lele
Cao, Qingwei Lin, et al. Thread: A logic-based data organization paradigm for how-to question
answering with retrieval augmented generation. arXiv preprint arXiv:2406.13372, 2024.
[29] Kaikai An, Fangkai Yang, Junting Lu, Liqun Li, Zhixing Ren, Hao Huang, Lu Wang, Pu Zhao, Yu Kang,
Hua Ding, et al. Nissist: An incident mitigation copilot based on troubleshooting guides. In
Proceedings of the 27th European Conference on Artificial Intelligence (ECAI 2024), pages 4471–4474,
2024.
[30] Kaikai An, Li Sheng, Ganqu Cui, Shuzheng Si, Ning Ding, Yu Cheng, and Baobao Chang. Ultraif:
Advancing instruction following from the wild. pages 7930–7957, 2025.
[31] Sumin An, Junyoung Sung, Wonpyo Park, Chanjun Park, and Paul Hongsuck Seo. Lcirc: A recurrent
compression approach for efficient long-form context and query dependent modeling in llms. North
American Chapter of the Association for Computational Linguistics, 2025.
[32] Sotiris Anagnostidis, Dario Pavllo, Luca Biggio, Lorenzo Noci, Aurélien Lucchi, and Thomas Hof-
mann. Dynamic context pruning for efficient and interpretable autoregressive transformers. Neural
Information Processing Systems, 2023.
[33] John R. Anderson, M. Matessa, and C. Lebiere. Act-r: A theory of higher level cognition and its
relation to visual attention. Hum. Comput. Interact., 1997.
[34] Jacob Andreas. Language models as agent models. Conference on Empirical Methods in Natural
Language Processing, 2022.
[35] Leonardo Aniello, R. Baldoni, and Leonardo Querzoni. Adaptive online scheduling in storm. Dis-
tributed Event-Based Systems, 2013.
[36] Petr Anokhin, Nikita Semenov, Artyom Sorokin, Dmitry Evseev, M. Burtsev, and Evgeny Burnaev.
Arigraph: Learning knowledge graph world models with episodic memory for llm agents, arXiv
preprint arXiv:2407.04363, 2024. URL https://arxiv.org/abs/2407.04363v3.
[37] Anthropic. Introducing the model context protocol, November 2024. URL https://www.
anthropic.com/news/model-context-protocol. [Online; accessed 17-July-2025].
[38] RM Aratchige and Dr. Wmks Ilmini. Llms working in harmony: A survey on the technological aspects
of building effective llm-based multi agent systems, arXiv preprint arXiv:2504.01963, 2025. URL
https://arxiv.org/abs/2504.01963v1.
[39] Leo Ardon, Daniel Furelos-Blanco, and A. Russo. Learning reward machines in cooperative multi-
agent tasks. AAMAS Workshops, 2023.
[40] K. Armeni, C. Honey, and Tal Linzen. Characterizing verbatim short-term memory in neural language
models. Conference on Computational Natural Language Learning, 2022.
[41] Akari Asai, Zeqiu Wu, Yizhong Wang, Avirup Sil, and Hannaneh Hajishirzi. Self-rag: Learning
to retrieve, generate, and critique through self-reflection. International Conference on Learning
Representations, 2023.
[42] Hikaru Asano, Tadashi Kozuno, and Yukino Baba. Self iterative label refinement via robust unla-
beled learning, arXiv preprint arXiv:2502.12565, 2025. URL https://arxiv.org/abs/2502.
12565v1.
[43] Ben Athiwaratkun, Sujan Kumar Gonugondla, Sanjay Krishna Gouda, Haifeng Qian, Hantian Ding,
Qing Sun, Jun Wang, Jiacheng Guo, Liangfu Chen, Parminder Bhatia, Ramesh Nallapati, Sudipta
Sengupta, and Bing Xiang. Bifurcated attention: Accelerating massively parallel decoding with
shared prefixes in llms, arXiv preprint arXiv:2403.08845, 2024. URL https://arxiv.org/abs/
2403.08845v2.
[44] Avinash Ayalasomayajula, Rui Guo, Jingbo Zhou, Sujan Kumar Saha, and Farimah Farahmandi. Lasp:
Llm assisted security property generation for soc verification. Workshop on Machine Learning for
CAD, 2024.
[45] Simon A. Aytes, Jinheon Baek, and Sung Ju Hwang. Sketch-of-thought: Efficient llm reasoning with
adaptive cognitive-inspired sketching. arXiv preprint, 2025.
[46] Bobby Azad, Reza Azad, Sania Eskandari, Afshin Bozorgpour, A. Kazerouni, I. Rekik, and D. Merhof.
Foundational models in medical imaging: A comprehensive survey and future vision, arXiv preprint
arXiv:2310.18689, 2023. URL https://arxiv.org/abs/2310.18689v1.
[47] Gilbert Badaro, Mohammed Saeed, and Paolo Papotti. Transformers for tabular data representation:
A survey of models and applications. Transactions of the Association for Computational Linguistics,
2023.
[48] Jinheon Baek, N. Chandrasekaran, Silviu Cucerzan, Allen Herring, and S. Jauhar. Knowledge-
augmented large language models for personalized contextual query suggestion. The Web Conference,
2023.
[49] Tianyi Bai, Hao Liang, Binwang Wan, Ling Yang, Bozhou Li, Yifan Wang, Bin Cui, Conghui He,
Binhang Yuan, and Wentao Zhang. A survey of multimodal large language model from a data-centric
perspective, arXiv preprint arXiv:2405.16640v2, 2024. URL https://arxiv.org/abs/2405.
16640v2.
[50] Yu Bai, Xiyuan Zou, Heyan Huang, Sanxing Chen, Marc-Antoine Rondeau, Yang Gao, and Jackie
Chi Kit Cheung. Citrus: Chunked instruction-aware state eviction for long sequence modeling.
Conference on Empirical Methods in Natural Language Processing, 2024.
[51] Yuntao Bai, Saurav Kadavath, Sandipan Kundu, Amanda Askell, Jackson Kernion, Andy Jones,
Anna Chen, Anna Goldie, Azalia Mirhoseini, Cameron McKinnon, Carol Chen, Catherine Olsson,
Christopher Olah, Danny Hernandez, Dawn Drain, Deep Ganguli, Dustin Li, Eli Tran-Johnson,
Ethan Perez, Jamie Kerr, Jared Mueller, Jeffrey Ladish, Joshua Landau, Kamal Ndousse, Kamile
Lukosuite, Liane Lovitt, Michael Sellitto, Nelson Elhage, Nicholas Schiefer, Noemi Mercado, Nova
DasSarma, Robert Lasenby, Robin Larson, Sam Ringer, Scott Johnston, Shauna Kravec, Sheer El
Showk, Stanislav Fort, Tamera Lanham, Timothy Telleen-Lawton, Tom Conerly, Tom Henighan,
Tristan Hume, Samuel R. Bowman, Zac Hatfield-Dodds, Ben Mann, Dario Amodei, Nicholas Joseph,
Sam McCandlish, Tom Brown, and Jared Kaplan. Constitutional ai: Harmlessness from ai feedback,
arXiv preprint arXiv:2212.08073, 2022. URL https://arxiv.org/abs/2212.08073.
[52] Souhail Bakkali, Sanket Biswas, Zuheng Ming, Mickaël Coustaty, Marccal Rusinol, O. R. Terrades,
and Josep Llad’os. Globaldoc: A cross-modal vision-language framework for real-world document
image retrieval and classification. IEEE Workshop/Winter Conference on Applications of Computer
Vision, 2023.
[53] Jayachandu Bandlamudi, K. Mukherjee, Prerna Agarwal, Sampath Dechu, Siyu Huo, Vatche Isahagian,
Vinod Muthusamy, N. Purushothaman, and Renuka Sindhgatta. Towards hybrid automation by
bootstrapping conversational interfaces for it operation tasks. AAAI Conference on Artificial Intelligence,
2023.
[54] Jayachandu Bandlamudi, Kushal Mukherjee, Prerna Agarwal, Ritwik Chaudhuri, R. Pimplikar,
Sampath Dechu, Alex Straley, Anbumunee Ponniah, and Renuka Sindhgatta. Building conversational
artifacts to enable digital assistant for apis and rpas. AAAI Conference on Artificial Intelligence, 2024.
[55] Keqin Bao, Jizhi Zhang, Xinyu Lin, Yang Zhang, Wenjie Wang, and Fuli Feng. Large language models
for recommendation: Past, present, and future. Annual International ACM SIGIR Conference on
Research and Development in Information Retrieval, 2024.
[56] Sara Di Bartolomeo, Giorgio Severi, V. Schetinger, and Cody Dunne. Ask and you shall receive
(a graph drawing): Testing chatgpt’s potential to apply graph layout algorithms. Eurographics
Conference on Visualization, 2023.
[57] Saikat Barua. Exploring autonomous agents through the lens of large language models: A review,
arXiv preprint arXiv:2404.04442, 2024. URL https://arxiv.org/abs/2404.04442v1.
[58] Kinjal Basu, Ibrahim Abdelaziz, Kelsey Bradford, M. Crouse, Kiran Kate, Sadhana Kumaravel, Saurabh
Goyal, Asim Munawar, Yara Rizk, Xin Wang, Luis A. Lastras, and P. Kapanipathi. Nestful: A benchmark
for evaluating llms on nested sequences of api calls, arXiv preprint arXiv:2409.03797, 2024. URL
https://arxiv.org/abs/2409.03797v3.
[59] Amin Beheshti. Natural language-oriented programming (nlop): Towards democratizing software
creation. 2024 IEEE International Conference on Software Services Engineering (SSE), 2024.
[60] Azadeh Beiranvand and S. M. Vahidipour. Integrating structural and semantic signals in text-
attributed graphs with bigtex, arXiv preprint arXiv:2504.12474, 2025. URL https://arxiv.org/
abs/2504.12474v2.
[61] Assaf Ben-Kish, Itamar Zimerman, Shady Abu-Hussein, Nadav Cohen, Amir Globerson, Lior Wolf,
and Raja Giryes. Decimamba: Exploring the length extrapolation potential of mamba. International
Conference on Learning Representations, 2024.
[62] Assaf Ben-Kish, Itamar Zimerman, M. J. Mirza, James R. Glass, Leonid Karlinsky, and Raja Giryes.
Overflow prevention enhances long-context recurrent llms. arXiv preprint, 2025.
[63] M. Benna and Stefano Fusi. Complex synapses as efficient memory systems. BMC Neuroscience,
2015.
[64] M. Benna and Stefano Fusi. Computational principles of biological memory, arXiv preprint
arXiv:1507.07580, 2015. URL https://arxiv.org/abs/1507.07580v1.
[65] Shelly Bensal, Umar Jamil, Christopher Bryant, M. Russak, Kiran Kamble, Dmytro Mozolevskyi,
Muayad Ali, and Waseem Alshikh. Reflect, retry, reward: Self-improving llms via reinforcement learn-
ing, arXiv preprint arXiv:2505.24726, 2025. URL https://arxiv.org/abs/2505.24726v1.
[66] Idoia Berges, J. Bermúdez, A. Goñi, and A. Illarramendi. Semantic web technology for agent
communication protocols. Extended Semantic Web Conference, 2008.
[67] Gaurav Beri and Vaishnavi Srivastava. Advanced techniques in prompt engineering for large language
models: A comprehensive study. 2024 IEEE 4th International Conference on ICT in Business Industry
& Government (ICTBIG), 2024.
[68] Amanda Bertsch, Uri Alon, Graham Neubig, and Matthew R. Gormley. Unlimiformer: Long-range
transformers with unlimited length input. Neural Information Processing Systems, 2023.
[69] Maciej Besta, Nils Blach, Aleš Kubíček, Robert Gerstenberger, Lukas Gianinazzi, Joanna Gajda,
Tomasz Lehmann, Michal Podstawski, H. Niewiadomski, P. Nyczyk, and Torsten Hoefler. Graph of
thoughts: Solving elaborate problems with large language models. AAAI Conference on Artificial
Intelligence, 2023.
[70] Gregor Betz and Kyle Richardson. Judgment aggregation, discursive dilemma and reflective equilib-
rium: Neural language models as self-improving doxastic agents. Frontiers in Artificial Intelligence,
2022.
[71] L. Bezalel, Eyal Orgad, and Amir Globerson. Teaching models to improve on tape. AAAI Conference
on Artificial Intelligence, 2024.
[72] Umang Bhatt, Sanyam Kapoor, Mihir Upadhyay, Ilia Sucholutsky, Francesco Quinzan, Katherine M.
Collins, Adrian Weller, Andrew Gordon Wilson, and Muhammad Bilal Zafar. When should we
orchestrate multiple agents?, arXiv preprint arXiv:2503.13577, 2025. URL https://arxiv.org/
abs/2503.13577v1.
[73] Baolong Bi, Shaohan Huang, Yiwei Wang, Tianchi Yang, Zihan Zhang, Haizhen Huang, Lingrui
Mei, Junfeng Fang, Zehao Li, Furu Wei, et al. Context-dpo: Aligning language models for context-
faithfulness. ACL 2025, 2024.
[74] Baolong Bi, Shenghua Liu, Lingrui Mei, Yiwei Wang, Pengliang Ji, and Xueqi Cheng. Decoding by
contrasting knowledge: Enhancing llms’ confidence on edited facts. ACL 2025, 2024.
[75] Baolong Bi, Shenghua Liu, Yiwei Wang, Lingrui Mei, and Xueqi Cheng. Lpnl: Scalable link prediction
with large language models. ACL 2024, 2024.
[76] Baolong Bi, Shenghua Liu, Yiwei Wang, Lingrui Mei, Hongcheng Gao, Junfeng Fang, and Xueqi
Cheng. Struedit: Structured outputs enable the fast and accurate knowledge editing for large
language models. 2024.
[77] Baolong Bi, Shenghua Liu, Yiwei Wang, Lingrui Mei, Hongcheng Gao, Yilong Xu, and Xueqi Cheng.
Adaptive token biaser: Knowledge editing via biasing key entities. EMNLP 2024, 2024.
[78] Baolong Bi, Shenghua Liu, Xingzhang Ren, Dayiheng Liu, Junyang Lin, Yiwei Wang, Lingrui Mei,
Junfeng Fang, Jiafeng Guo, and Xueqi Cheng. Refinex: Learning to refine pre-training data at scale
from expert-guided programs. 2025.
[79] Baolong Bi, Shenghua Liu, Yiwei Wang, Lingrui Mei, Junfeng Fang, Hongcheng Gao, Shiyu Ni, and
Xueqi Cheng. Is factuality enhancement a free lunch for llms? better factuality can lead to worse
context-faithfulness. ICLR 2025, 2025.
[80] Baolong Bi, Shenghua Liu, Yiwei Wang, Yilong Xu, Junfeng Fang, Lingrui Mei, and Xueqi Cheng.
Parameters vs. context: Fine-grained control of knowledge reliance in language models. 2025.
[81] Bin Bi, Chenliang Li, Chen Wu, Ming Yan, and Wei Wang. Palm: Pre-training an autoencod-
ing&autoregressive language model for context-conditioned generation. Conference on Empirical
Methods in Natural Language Processing, 2020.
[82] Dinh Doan Van Bien, David Lillis, and Rem W. Collier. Call graph profiling for multi agent systems.
Multi-Agent Logics, Languages, and Organisations Federated Workshops, 2009.
[83] Jonas Bode, Bastian Pätzold, Raphael Memmesheimer, and Sven Behnke. A comparison of prompt
engineering techniques for task planning and execution in service robotics. IEEE-RAS International
Conference on Humanoid Robots, 2024.
[84] P. Bonzon. Grounding mental representations in a virtual multi-level functional framework. Journal
of Cognition, 2023.
[85] Sebastian Borgeaud, A. Mensch, Jordan Hoffmann, Trevor Cai, Eliza Rutherford, Katie Millican,
George van den Driessche, Jean-Baptiste Lespiau, Bogdan Damoc, Aidan Clark, Diego de Las Casas,
Aurelia Guy, Jacob Menick, Roman Ring, T. Hennigan, Saffron Huang, Lorenzo Maggiore, Chris Jones,
Albin Cassirer, Andy Brock, Michela Paganini, G. Irving, O. Vinyals, Simon Osindero, K. Simonyan,
Jack W. Rae, Erich Elsen, and L. Sifre. Improving language models by retrieving from trillions of
tokens. International Conference on Machine Learning, 2021.
[86] Zalán Borsos, Raphaël Marinier, Damien Vincent, E. Kharitonov, O. Pietquin, Matthew Sharifi,
Dominik Roblek, O. Teboul, David Grangier, M. Tagliasacchi, and Neil Zeghidour. Audiolm: A
language modeling approach to audio generation. IEEE/ACM Transactions on Audio Speech and
Language Processing, 2022.
[87] FutureSearch Nikos I. Bosse, Jon Evans, Robert G. Gambee, Daniel Hnyk, Peter Muhlbacher, Lawrence
Phillips, Dan Schwarz, and Jack Wildman. Deep research bench: Evaluating ai web research agents,
arXiv preprint arXiv:2506.06287, 2025. URL https://arxiv.org/abs/2506.06287v1.
[88] Vicent Botti. Agentic ai and multiagentic: Are we reinventing the wheel?, arXiv preprint
arXiv:2506.01463, 2025. URL https://arxiv.org/abs/2506.01463v1.
[89] William Brach, Kristián Kostál, and Michal Ries. The effectiveness of large language models in
transforming unstructured text to standardized formats. IEEE Access, 2025.
[90] C. Brainerd, C. F. Gomes, and K. Nakamura. Dual recollection in episodic memory. Journal of
experimental psychology. General, 2015.
[91] Inês Bramão, Jiefeng Jiang, A. Wagner, and M. Johansson. Encoding contexts are incidentally
reinstated during competitive retrieval and track the temporal dynamics of memory interference.
Cerebral Cortex, 2022.
[92] Andrés M Bran, Sam Cox, Oliver Schilter, Carlo Baldassari, Andrew D. White, and P. Schwaller.
Augmenting large language models with chemistry tools. Nat. Mac. Intell., 2023.
[93] Maricela Claudia Bravo and Martha Coronel. Aligning agent communication protocols - a pragmatic
approach. International Conference on Software and Data Technologies, 2008.
[94] F. Brazier, B. Dunin-Keplicz, N. Jennings, and Jan Treur. Desire: Modelling multi-agent systems in a
compositional formal framework. International Journal of Cooperative Information Systems, 1997.
[95] Lorenz Brehme, Thomas Ströhle, and Ruth Breu. Can llms be trusted for evaluating rag systems? a
survey of methods and datasets, arXiv preprint arXiv:2504.20119, 2025. URL https://arxiv.
org/abs/2504.20119v2.
[96] R. Breil, D. Delahaye, Laurent Lapasset, and E. Feron. Multi-agent systems to help managing air
traffic structure. arXiv preprint, 2017.
[97] Alexander Brinkmann and Christian Bizer. Self-refinement strategies for llm-based product attribute
value extraction. Datenbanksysteme für Business, Technologie und Web, 2025.
[98] D. Britz, M. Guan, and Minh-Thang Luong. Efficient attention using a fixed-size memory representa-
tion. Conference on Empirical Methods in Natural Language Processing, 2017.
[99] Adam W. Broitman and M. J. Kahana. Neural context reinstatement of recurring events. bioRxiv,
2024.
[100] Ethan A. Brooks, Logan Walls, Richard L. Lewis, and Satinder Singh. Large language models can
implement policy iteration. Neural Information Processing Systems, 2022.
[101] Rodney A. Brooks. A robust layered control system for a mobile robot. IEEE J. Robotics Autom., 1986.
[102] Tim Brooks, Aleksander Holynski, and Alexei A. Efros. Instructpix2pix: Learning to follow image
editing instructions. Computer Vision and Pattern Recognition, 2022.
[103] Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, J. Kaplan, Prafulla Dhariwal, Arvind
Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss,
Gretchen Krueger, T. Henighan, R. Child, A. Ramesh, Daniel M. Ziegler, Jeff Wu, Clemens Winter,
Christopher Hesse, Mark Chen, Eric Sigler, Ma teusz Litwin, Scott Gray, Benjamin Chess, Jack Clark,
Christopher Berner, Sam McCandlish, Alec Radford, I. Sutskever, and Dario Amodei. Language
models are few-shot learners. Neural Information Processing Systems, 2020.
[104] Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal,
Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-
Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu,
Clemens Winter, Christopher Hesse, Mark Chen, Eric Sigler, Mateusz Litwin, Scott Gray, Benjamin
Chess, Jack Clark, Christopher Berner, Sam McCandlish, Alec Radford, Ilya Sutskever, and Dario
Amodei. Language models are few-shot learners, arXiv preprint arXiv:2005.14165, 2020. URL
https://arxiv.org/abs/2005.14165.
[105] Davide Caffagni, Federico Cocchi, Nicholas Moratelli, Sara Sarto, Marcella Cornia, L. Baraldi, and
R. Cucchiara. Wiki-llava: Hierarchical retrieval-augmented generation for multimodal llms. 2024
IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2024.
[106] Joyce Cahoon, Prerna Singh, Nick Litombe, Jonathan Larson, Ha Trinh, Yiwen Zhu, Andreas Mueller,
Fotis Psallidas, and Carlo Curino. Optimizing open-domain question answering with graph-based
retrieval augmented generation, arXiv preprint arXiv:2503.02922, 2025. URL https://arxiv.
org/abs/2503.02922v1.
[107] Hongru Cai, Yongqi Li, Wenjie Wang, Fengbin Zhu, Xiaoyu Shen, Wenjie Li, and Tat-Seng Chua.
Large language models empowered personalized web agents. The Web Conference, 2024.
[108] Yujun Cai, Liuhao Ge, Jianfei Cai, and Junsong Yuan. Weakly-supervised 3d hand pose estimation
from monocular rgb images. In Proceedings of the European conference on computer vision (ECCV),
pages 666–682, 2018.
[109] Yujun Cai, Liuhao Ge, Jun Liu, Jianfei Cai, Tat-Jen Cham, Junsong Yuan, and Nadia Magnenat
Thalmann. Exploiting spatial-temporal relationships for 3d pose estimation via graph convolutional
networks. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 2272–
2281, 2019.
[110] Yujun Cai, Liuhao Ge, Jianfei Cai, Nadia Magnenat Thalmann, and Junsong Yuan. 3d hand pose
estimation using synthetic data and weakly labeled rgb images. IEEE transactions on pattern analysis
and machine intelligence, 43(11):3739–3753, 2020.
[111] Yujun Cai, Lin Huang, Yiwei Wang, Tat-Jen Cham, Jianfei Cai, Junsong Yuan, Jun Liu, Xu Yang, Yiheng
Zhu, Xiaohui Shen, et al. Learning progressive joint propagation for human motion prediction.
In Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020,
Proceedings, Part VII 16, pages 226–242. Springer International Publishing, 2020.
[112] Yujun Cai, Yiwei Wang, Yiheng Zhu, Tat-Jen Cham, Jianfei Cai, Junsong Yuan, Jun Liu, Chuanxia
Zheng, Sijie Yan, Henghui Ding, et al. A unified 3d human motion synthesis model via conditional
variational auto-encoder. In Proceedings of the IEEE/CVF International Conference on Computer Vision,
pages 11645–11655, 2021.
[113] V. Camos and P. Barrouillet. Attentional and non-attentional systems in the maintenance of ver-
bal information in working memory: the executive and phonological loops. Frontiers in Human
Neuroscience, 2014.
[114] Boxi Cao, Qiaoyu Tang, Hongyu Lin, Xianpei Han, Jiawei Chen, Tianshu Wang, and Le Sun. Retentive
or forgetful? diving into the knowledge memorizing mechanism of language models. International
Conference on Language Resources and Evaluation, 2023.
[115] He Cao, Zhenwei An, Jiazhan Feng, Kun Xu, Liwei Chen, and Dongyan Zhao. A step closer to com-
prehensive answers: Constrained multi-stage question decomposition with large language models,
arXiv preprint arXiv:2311.07491, 2023. URL https://arxiv.org/abs/2311.07491v1.
[116] Nicola De Cao, Wilker Aziz, and Ivan Titov. Question answering by reasoning across documents
with graph convolutional networks. North American Chapter of the Association for Computational
Linguistics, 2018.
[117] Pengfei Cao, Tianyi Men, Wencan Liu, Jingwen Zhang, Xuzhao Li, Xixun Lin, Dianbo Sui, Yanan Cao,
Kang Liu, and Jun Zhao. Large language models for planning: A comprehensive and systematic survey,
arXiv preprint arXiv:2505.19683, 2025. URL https://arxiv.org/abs/2505.19683v1.
[118] Yongcan Cao, Wenwu Yu, W. Ren, and Guanrong Chen. An overview of recent progress in the study
of distributed multi-agent coordination. IEEE Transactions on Industrial Informatics, 2012.
[119] Yuji Cao, Huan Zhao, Yuheng Cheng, Ting Shu, Yue Chen, Guolong Liu, Gaoqi Liang, Junhua
Zhao, Jinyue Yan, and Yunjie Li. Survey on large language model-enhanced reinforcement learning:
Concept, taxonomy, and methods. IEEE Transactions on Neural Networks and Learning Systems, 2024.
[120] Yukun Cao, Zengyi Gao, Zhiyang Li, Xike Xie, and S. K. Zhou. Lego-graphrag: Modularizing
graph-based retrieval-augmented generation for design space exploration. arXiv preprint, 2024.
[121] R. C. Cardoso and Angelo Ferrando. A review of agent-based programming for multi-agent systems.
De Computis, 2021.
[122] Nicholas Carlini, Chang Liu, Ú. Erlingsson, Jernej Kos, and D. Song. The secret sharer: Evaluating
and testing unintended memorization in neural networks. USENIX Security Symposium, 2018.
[123] Nicholas Carlini, Florian Tramèr, Eric Wallace, Matthew Jagielski, Ariel Herbert-Voss, Katherine Lee,
Adam Roberts, Tom B. Brown, D. Song, Ú. Erlingsson, Alina Oprea, and Colin Raffel. Extracting
training data from large language models. USENIX Security Symposium, 2020.
[124] Daniel Casanueva-Morato, A. Ayuso-Martinez, J. P. Dominguez-Morales, A. Jiménez-Fernandez, and
G. Jiménez-Moreno. A bio-inspired implementation of a sparse-learning spike-based hippocampus
memory model. IEEE Transactions on Emerging Topics in Computing, 2022.
[125] Daniel Casanueva-Morato, A. Ayuso-Martinez, J. P. Dominguez-Morales, A. Jiménez-Fernandez, and
G. Jiménez-Moreno. Bio-inspired computational memory model of the hippocampus: an approach
to a neuromorphic spike-based content-addressable memory. Neural Networks, 2023.
[126] Amartya Chakraborty, Paresh Dashore, Nadia Bathaee, Anmol Jain, Anirban Das, Shi-Xiong Zhang,
Sambit Sahu, M. Naphade, and Genta Indra Winata. T1: A tool-oriented conversational dataset for
multi-turn agentic planning, arXiv preprint arXiv:2505.16986, 2025. URL https://arxiv.org/
abs/2505.16986v1.
[127] Kranti Chalamalasetti, Jana Gotze, Sherzod Hakimov, Brielen Madureira, Philipp Sadler, and David
Schlangen. clembench: Using game play to evaluate chat-optimized language models as conversa-
tional agents. Conference on Empirical Methods in Natural Language Processing, 2023.
[128] Edward Y. Chang and Longling Geng. Sagallm: Context management, validation, and transaction
guarantees for multi-agent llm planning, arXiv preprint arXiv:2503.11951, 2025. URL https:
//arxiv.org/abs/2503.11951v2.
[129] Yu-Chu Chang, Xu Wang, Jindong Wang, Yuan Wu, Kaijie Zhu, Hao Chen, Linyi Yang, Xiaoyuan
Yi, Cunxiang Wang, Yidong Wang, Weirong Ye, Yue Zhang, Yi Chang, Philip S. Yu, Qian Yang, and
Xingxu Xie. A survey on evaluation of large language models. ACM Transactions on Intelligent Systems
and Technology, 2023.
[130] Subhajit Chaudhury, Payel Das, Sarathkrishna Swaminathan, Georgios Kollias, Elliot Nelson, Khushbu
Pahwa, Tejaswini Pedapati, Igor Melnyk, and Matthew Riemer. Epman: Episodic memory attention
for generalizing to longer contexts, arXiv preprint arXiv:2502.14280, 2025. URL https://arxiv.
org/abs/2502.14280v1.
[131] Xueqi CHEGN, Shenghua Liu, and Ruqing ZHANG. Thinking on new system for big data technology.
Bulletin of Chinese Academy of Sciences (Chinese Version), 37(1):60–67, 2022.
[132] Viktoriia Chekalina, Anton Razzigaev, Elizaveta Goncharova, and Andrey Kuznetsov. Addressing
hallucinations in language models with knowledge graph embeddings as an additional modality,
arXiv preprint arXiv:2411.11531, 2024. URL https://arxiv.org/abs/2411.11531v2.
[133] Bo Chen, Yingyu Liang, Zhizhou Sha, Zhenmei Shi, and Zhao Song. Hsr-enhanced sparse attention
acceleration, arXiv preprint arXiv:2410.10165, 2024. URL https://arxiv.org/abs/2410.
10165v2.
[134] Boyu Chen, Zirui Guo, Zidan Yang, Yuluo Chen, Junze Chen, Zhenghao Liu, Chuan Shi, and Cheng
Yang. Pathrag: Pruning graph-based retrieval augmented generation with relational paths, arXiv
preprint arXiv:2502.14902, 2025. URL https://arxiv.org/abs/2502.14902v1.
[135] Fei-Long Chen, Du-Zhen Zhang, Ming-Lun Han, Xiu-Yi Chen, Jing Shi, Shuang Xu, and Bo Xu. Vlp:
A survey on vision-language pre-training. Machine Intelligence Research, 20(1):38–56, 2023.
[136] Feiyang Chen, Yu Cheng, Lei Wang, Yuqing Xia, Ziming Miao, Lingxiao Ma, Fan Yang, Jilong
Xue, Zhi Yang, Mao Yang, and Haibo Chen. Attentionengine: A versatile framework for efficient
attention mechanisms on diverse hardware platforms, arXiv preprint arXiv:2502.15349, 2025. URL
https://arxiv.org/abs/2502.15349v1.
[137] Huajun Chen. Large knowledge model: Perspectives and challenges. Data Intelligence, 2023.
[138] Jianing Chen, Zehao Li, Yujun Cai, Hao Jiang, Chengxuan Qian, Juyuan Kang, Shuqin Gao, Honglong
Zhao, Tianlu Mao, and Yucheng Zhang. Haif-gs: Hierarchical and induced flow-guided gaussian
splatting for dynamic scene. 2025.
[139] Jiaqi Chen, Xiaoye Zhu, Yue Wang, Tianyang Liu, Xinhui Chen, Ying Chen, Chak Tou Leong, Yifei
Ke, Joseph Liu, Yiwen Yuan, Julian McAuley, and Li jia Li. Symbolic representation for any-to-any
generative tasks, arXiv preprint arXiv:2504.17261v1, 2025. URL https://arxiv.org/abs/
2504.17261v1.
[140] Jiayi Chen, J. Ye, and Guiling Wang. From standalone llms to integrated intelligence: A survey of
compound al systems, arXiv preprint arXiv:2506.04565, 2025. URL https://arxiv.org/abs/
2506.04565v1.
[141] Jin Chen, Zheng Liu, Xu Huang, Chenwang Wu, Qi Liu, Gangwei Jiang, Yuanhao Pu, Yuxuan Lei,
Xiaolong Chen, Xingmei Wang, Defu Lian, and Enhong Chen. When large language models meet
personalization: Perspectives of challenges and opportunities. World wide web (Bussum), 2023.
[142] Jiyu Chen, Shuang Peng, Daxiong Luo, Fan Yang, Renshou Wu, Fangyuan Li, and Xiaoxin Chen.
Edgeinfinite: A memory-efficient infinite-context transformer for edge devices, arXiv preprint
arXiv:2503.22196, 2025. URL https://arxiv.org/abs/2503.22196v1.
[143] Justin Chih-Yao Chen, Archiki Prasad, Swarnadeep Saha, Elias Stengel-Eskin, and Mohit Bansal. Magi-
core: Multi-agent, iterative, coarse-to-fine refinement for reasoning, arXiv preprint arXiv:2409.12147,
2024. URL https://arxiv.org/abs/2409.12147v1.
[144] Mingyang Chen, Haoze Sun, Tianpeng Li, Fan Yang, Hao Liang, Keer Lu, Bin Cui, Wentao Zhang,
Zenan Zhou, and Weipeng Chen. Facilitating multi-turn function calling for llms via compositional
instruction tuning. International Conference on Learning Representations, 2024.
[145] Nuo Chen, Yuhan Li, Jianheng Tang, and Jia Li. Graphwiz: An instruction-following language model
for graph computational problems. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge
Discovery and Data Mining, pages 353–364, 2024.
[146] Nuo Chen, Zhiyuan Hu, Qingyun Zou, Jiaying Wu, Qian Wang, Bryan Hooi, and Bingsheng He.
Judgelrm: Large reasoning models as a judge, arXiv preprint arXiv:2504.00050, 2025. URL https:
//arxiv.org/abs/2504.00050v1.
[147] Qiguang Chen, Libo Qin, Jiaqi Wang, Jinxuan Zhou, and Wanxiang Che. Unlocking the capabilities
of thought: A reasoning boundary framework to quantify and optimize chain-of-thought, arXiv
preprint arXiv:2410.05695, 2024. URL https://arxiv.org/abs/2410.05695.
[148] Qiguang Chen, Libo Qin, Jinhao Liu, Dengyun Peng, Jiannan Guan, Peng Wang, Mengkang Hu,
Yuhang Zhou, Te Gao, and Wangxiang Che. Towards reasoning era: A survey of long chain-
of-thought for reasoning large language models, arXiv preprint arXiv:2503.09567, 2025. URL
https://arxiv.org/abs/2503.09567v3.
[149] Qiguang Chen, Libo Qin, Jinhao Liu, Dengyun Peng, Jiannan Guan, Peng Wang, Mengkang Hu,
Yuhang Zhou, Te Gao, and Wanxiang Che. Towards reasoning era: A survey of long chain-of-
thought for reasoning large language models, arXiv preprint arXiv:2503.09567, 2025. URL https:
//arxiv.org/abs/2503.09567.
[150] Qiguang Chen, Libo Qin, Jinhao Liu, Dengyun Peng, Jiaqi Wang, Mengkang Hu, Zhi Chen, Wanxiang
Che, and Ting Liu. Ecm: A unified electronic circuit model for explaining the emergence of in-context
learning and chain-of-thought in large language model, arXiv preprint arXiv:2502.03325, 2025.
URL https://arxiv.org/abs/2502.03325.
[151] Qiguang Chen, Mingda Yang, Libo Qin, Jinhao Liu, Zheng Yan, Jiannan Guan, Dengyun Peng, Yiyan
Ji, Hanjing Li, Mengkang Hu, Yimeng Zhang, Yihao Liang, Yuhang Zhou, Jiaqi Wang, Zhi Chen, and
Wanxiang Che. Ai4research: A survey of artificial intelligence for scientific research, arXiv preprint
arXiv:2507.01903, 2025. URL https://arxiv.org/abs/2507.01903.
[152] S Chen, Y Wang, YF Wu, and Q Chen. . . . Advancing tool-augmented large
language models: Integrating insights from errors in inference trees. 2024.
URL https://proceedings.neurips.cc/paper_files/paper/2024/hash/
c0f7ee1901fef1da4dae2b88dfd43195-Abstract-Conference.html.
[153] Shouyuan Chen, Sherman Wong, Liangjian Chen, and Yuandong Tian. Extending context window
of large language models via positional interpolation, arXiv preprint arXiv:2306.15595, 2023. URL
https://arxiv.org/abs/2306.15595v2.
[154] Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey E. Hinton. A simple framework for
contrastive learning of visual representations. International Conference on Machine Learning, 2020.
[155] Wenhu Chen, Xueguang Ma, Xinyi Wang, and William W. Cohen. Program of thoughts prompting:
Disentangling computation from reasoning for numerical reasoning tasks. Trans. Mach. Learn. Res.,
2022.
[156] Yanda Chen, Ruiqi Zhong, Sheng Zha, G. Karypis, and He He. Meta-learning via language model
in-context tuning. Annual Meeting of the Association for Computational Linguistics, 2021.
[157] Yi Chen, JiaHao Zhao, and HaoHao Han. A survey on collaborative mechanisms between large and
small language models, arXiv preprint arXiv:2505.07460, 2025. URL https://arxiv.org/abs/
2505.07460v1.
[158] Yixin Chen, Shuai Zhang, Boran Han, Tong He, and Bo Li. Camml: Context-aware multimodal
learner for large models. Annual Meeting of the Association for Computational Linguistics, 2024.
[159] Z Chen, K Zhou, B Zhang, Z Gong, and WX Zhao. . . . Chatcot: Tool-augmented chain-of-thought
reasoning on chat-based large language models. 2023. URL https://arxiv.org/abs/2305.
14323.
[160] Zehui Chen, Weihua Du, Wenwei Zhang, Kuikun Liu, Jiangning Liu, Miao Zheng, Jingming Zhuo,
Songyang Zhang, Dahua Lin, Kai Chen, et al. T-eval: Evaluating the tool utilization capability step
by step. arXiv preprint arXiv:2312.14033, 2023.
[161] Zehui Chen, Kuikun Liu, Qiuchen Wang, Jiangning Liu, Wenwei Zhang, Kai Chen, and Feng Zhao.
Mindsearch: Mimicking human minds elicits deep ai searcher, arXiv preprint arXiv:2407.20183,
2024. URL https://arxiv.org/abs/2407.20183v1.
[162] Zhi Chen, Qiguang Chen, Libo Qin, Qipeng Guo, Haijun Lv, Yicheng Zou, Wanxiang Che, Hang
Yan, Kai Chen, and Dahua Lin. What are the essential factors in crafting effective long context
multi-hop instruction datasets? insights and best practices, arXiv preprint arXiv:2409.01893, 2025.
URL https://arxiv.org/abs/2409.01893.
[163] Zhikai Chen, Haitao Mao, Hang Li, Wei Jin, Haifang Wen, Xiaochi Wei, Shuaiqiang Wang, Dawei
Yin, Wenqi Fan, Hui Liu, and Jiliang Tang. Exploring the potential of large language models (llms)in
learning on graphs. SIGKDD Explorations, 2023.
[164] Zihan Chen, Song Wang, Zhen Tan, Xingbo Fu, Zhenyu Lei, Peng Wang, Huan Liu, Cong Shen, and
Jundong Li. A survey of scaling in large language model reasoning, arXiv preprint arXiv:2504.02181,
2025. URL https://arxiv.org/abs/2504.02181v1.
[165] ZY Chen, S Shen, G Shen, and G Zhi. . . . Towards tool use alignment of large language models.
2024. URL https://aclanthology.org/2024.emnlp-main.82/.
[166] Mingyue Cheng, Yucong Luo, Ouyang Jie, Qi Liu, Huijie Liu, Li Li, Shuo Yu, Bohou Zhang, Jiawei Cao,
Jie Ma, Daoyu Wang, and Enhong Chen. A survey on knowledge-oriented retrieval-augmented gener-
ation, arXiv preprint arXiv:2503.10677, 2025. URL https://arxiv.org/abs/2503.10677v2.
[167] Ning Cheng, Zhaohui Yan, Ziming Wang, Zhijie Li, Jiaming Yu, Zilong Zheng, Kewei Tu, Jinan Xu,
and Wenjuan Han. Potential and limitations of llms in capturing structured semantics: A case study
on srl. International Conference on Intelligent Computing, 2024.
[168] Sitao Cheng, Ziyuan Zhuang, Yong Xu, Fangkai Yang, Chaoyun Zhang, Xiaoting Qin, Xiang Huang,
Ling Chen, Qingwei Lin, Dongmei Zhang, et al. Call me when necessary: Llms can efficiently and
faithfully reason over structured environments. In Association for Computational Linguistics 2024,
pages 4275–4295, 2024.
[169] Xin Cheng, Di Luo, Xiuying Chen, Lemao Liu, Dongyan Zhao, and Rui Yan. Lift yourself up:
Retrieval-augmented text generation with self memory. Neural Information Processing Systems, 2023.
[170] Yao Cheng, Yibo Zhao, Jiapeng Zhu, Yao Liu, Xing Sun, and Xiang Li. Human cognition inspired rag
with knowledge graph for complex problem solving, arXiv preprint arXiv:2503.06567, 2025. URL
https://arxiv.org/abs/2503.06567v1.
[171] Yuheng Cheng, Ceyao Zhang, Zhengwen Zhang, Xiangrui Meng, Sirui Hong, Wenhao Li, Zihao
Wang, Zekai Wang, Feng Yin, Junhua Zhao, and Xiuqiang He. Exploring large language model based
intelligent agents: Definitions, methods, and prospects, arXiv preprint arXiv:2401.03428, 2024. URL
https://arxiv.org/abs/2401.03428v1.
[172] Egor Cherepanov, Nikita Kachaev, A. Kovalev, and Aleksandr I. Panov. Memory, benchmark & robots: A
benchmark for solving complex tasks with reinforcement learning, arXiv preprint arXiv:2502.10550,
2025. URL https://arxiv.org/abs/2502.10550v2.
[173] Prateek Chhikara, Dev Khant, Saket Aryan, Taranjeet Singh, and Deshraj Yadav. Mem0: Building
production-ready ai agents with scalable long-term memory, arXiv preprint arXiv:2504.19413, 2025.
URL https://arxiv.org/abs/2504.19413.
[174] Yew Ken Chia, Lidong Bing, Soujanya Poria, and Luo Si. Relationprompt: Leveraging prompts to
generate synthetic data for zero-shot relation triplet extraction. Findings, 2022.
[175] Jihye Choi, Nils Palumbo, P. Chalasani, Matthew M. Engelhard, Somesh Jha, Anivarya Kumar, and
David Page. Malade: Orchestration of llm-powered agents with retrieval augmented generation
for pharmacovigilance, arXiv preprint arXiv:2408.01869, 2024. URL https://arxiv.org/abs/
2408.01869v1.
[176] K. Choromanski, Valerii Likhosherstov, David Dohan, Xingyou Song, Andreea Gane, Tamás Sarlós,
Peter Hawkins, Jared Davis, Afroz Mohiuddin, Lukasz Kaiser, David Belanger, Lucy J. Colwell,
and Adrian Weller. Rethinking attention with performers. International Conference on Learning
Representations, 2020.
[177] Zhendong Chu, Shen Wang, Jian Xie, Tinghui Zhu, Yibo Yan, Jinheng Ye, Aoxiao Zhong, Xuming Hu,
Jing Liang, Philip S. Yu, and Qingsong Wen. Llm agents for education: Advances and applications,
arXiv preprint arXiv:2503.11733, 2025. URL https://arxiv.org/abs/2503.11733v1.
[178] Zhixuan Chu, Huaiyu Guo, Xinyuan Zhou, Yijia Wang, Fei Yu, Hong Chen, Wanqing Xu, Xin Lu, Qing
Cui, Longfei Li, Junqing Zhou, and Sheng Li. Data-centric financial large language models, arXiv
preprint arXiv:2310.17784, 2023. URL https://arxiv.org/abs/2310.17784v2.
[179] Yun-Shiuan Chuang, Agam Goyal, Nikunj Harlalka, Siddharth Suresh, Robert Hawkins, Sijia Yang,
Dhavan Shah, Junjie Hu, and Timothy T. Rogers. Simulating opinion dynamics with networks of
llm-based agents, arXiv preprint arXiv:2311.09618, 2024. URL https://arxiv.org/abs/2311.
09618.
[180] Yung-Sung Chuang, Benjamin Cohen-Wang, Shannon Zejiang Shen, Zhaofeng Wu, Hu Xu, Xi Victoria
Lin, James Glass, Shang-Wen Li, and Wen tau Yih. Selfcite: Self-supervised alignment for context
attribution in large language models, arXiv preprint arXiv:2502.09604, 2025. URL https://
arxiv.org/abs/2502.09604v3.
[181] Julian Coda-Forno, Marcel Binz, Zeynep Akata, M. Botvinick, Jane X. Wang, and Eric Schulz.
Meta-in-context learning in large language models. Neural Information Processing Systems, 2023.
[182] Joao Coelho, Jingjie Ning, Jingyuan He, Kangrui Mao, A. Paladugu, Pranav Setlur, Jiahe Jin, James P.
Callan, João Magalhães, Bruno Martins, and Chenyan Xiong. Deepresearchgym: A free, transparent,
and reproducible evaluation sandbox for deep research, arXiv preprint arXiv:2505.19253, 2025.
URL https://arxiv.org/abs/2505.19253v2.
[183] Emile Contal and Garrin McGoldrick. Ragsys: Item-cold-start recommender as rag system. IR-
RAG@SIGIR, 2024.
[184] Erica Coppolillo. Injecting knowledge graphs into large language models, arXiv preprint
arXiv:2505.07554, 2025. URL https://arxiv.org/abs/2505.07554v1.
[185] R. P. Costa, R. Froemke, P. J. Sjöström, and Mark C. W. van Rossum. Unified pre- and postsynaptic
long-term plasticity enables reliable and flexible learning. eLife, 2015.
[186] Caia Costello, Simon Guo, Anna Goldie, and Azalia Mirhoseini. Think, prune, train, improve:
Scaling reasoning without scaling models, arXiv preprint arXiv:2504.18116, 2025. URL https:
//arxiv.org/abs/2504.18116v1.
[187] Michael Craig, Karla Butterworth, Jonna Nilsson, Colin J Hamilton, P. Gallagher, and T. Smulders.
How does intentionality of encoding affect memory for episodic information? Learning & memory
(Cold Spring Harbor, N.Y.), 2016.
[188] crewAI Inc. crewai: Framework for orchestrating role-playing, autonomous ai agents. https:
//github.com/crewAIInc/crewAI, 2024. [Online; accessed 17-July-2025].
[189] A. Cruz, André V. dos Santos, R. Santiago, and B. Bedregal. A fuzzy semantic for bdi logic. Fuzzy
Information and Engineering, 2021.
[190] Florin Cuconasu, Giovanni Trappolini, F. Siciliano, Simone Filice, Cesare Campagnano, Y. Maarek,
Nicola Tonellotto, and Fabrizio Silvestri. The power of noise: Redefining retrieval for rag systems.
Annual International ACM SIGIR Conference on Research and Development in Information Retrieval,
2024.
[191] Kai Cui, Anam Tahir, Gizem Ekinci, Ahmed Elshamanhory, Yannick Eich, Mengguang Li, and
H. Koeppl. A survey on large-population systems and scalable multi-agent reinforcement learning,
arXiv preprint arXiv:2209.03859, 2022. URL https://arxiv.org/abs/2209.03859v1.
[192] Yuanning Cui, Zequn Sun, and Wei Hu. A prompt-based knowledge graph foundation model for
universal in-context reasoning. In Advances in Neural Information Processing Systems, 2024.
[193] Yue Cui, Liuyi Yao, Shuchang Tao, Weijie Shi, Yaliang Li, Bolin Ding, and Xiaofang Zhou. En-
hancing tool learning in large language models with hierarchical error checklists, arXiv preprint
arXiv:2506.00042, 2025. URL https://arxiv.org/abs/2506.00042v1.
[194] C. Curto, A. Degeratu, and V. Itskov. Flexible memory networks. Bulletin of Mathematical Biology,
2010.
[195] Ruiting Dai, Yuqiao Tan, Lisi Mo, Shuang Liang, Guohao Huo, Jiayi Luo, and Yao Cheng. G-sap:
Graph-based structure-aware prompt learning over heterogeneous knowledge for commonsense
reasoning. International Conference on Multimedia Retrieval, 2024.
[196] Xinnan Dai, Haohao Qu, Yifen Shen, Bohang Zhang, Qihao Wen, Wenqi Fan, Dongsheng Li, Jiliang
Tang, and Caihua Shan. How do large language models understand graph patterns? a benchmark for
graph pattern comprehension, arXiv preprint arXiv:2410.05298v2, 2024. URL https://arxiv.
org/abs/2410.05298v2.
[197] Fatemeh Daneshfar and H. Bevrani. Multi-agent systems in control engineering: a survey. arXiv
preprint, 2009.
[198] Yufan Dang, Cheng Qian, Xueheng Luo, Jingru Fan, Zihao Xie, Ruijie Shi, Weize Chen, Cheng
Yang, Xiaoyin Che, Ye Tian, Xuantang Xiong, Lei Han, Zhiyuan Liu, and Maosong Sun. Multi-agent
collaboration via evolving orchestration, arXiv preprint arXiv:2505.19591, 2025. URL https:
//arxiv.org/abs/2505.19591v1.
[199] Tri Dao. Flashattention-2: Faster attention with better parallelism and work partitioning. International
Conference on Learning Representations, 2023.
[200] Tri Dao, Daniel Y. Fu, Stefano Ermon, A. Rudra, and Christopher R’e. Flashattention: Fast and
memory-efficient exact attention with io-awareness. Neural Information Processing Systems, 2022.
[201] D Das, D Banerjee, S Aditya, and A Kulkarni. Mathsensei: a tool-augmented large language model
for mathematical reasoning. 2024. URL https://arxiv.org/abs/2402.17231.
[202] Payel Das, Subhajit Chaudhury, Elliot Nelson, Igor Melnyk, Sarath Swaminathan, Sihui Dai, Aurélie
Lozano, Georgios Kollias, Vijil Chenthamarakshan, Jiří, Navrátil, Soham Dan, and Pin-Yu Chen.
Larimar: Large language models with episodic memory control, arXiv preprint arXiv:2403.11901,
2024. URL https://arxiv.org/abs/2403.11901.
[203] Adrian de Wynter, Xun Wang, Qilong Gu, and Si-Qing Chen. On meta-prompting, arXiv preprint
arXiv:2312.06562, 2023. URL https://arxiv.org/abs/2312.06562v3.
[204] Ramandeep Singh Dehal, Mehak Sharma, and Enayat Rajabi. Knowledge graphs and their reciprocal
relationship with large language models. Machine Learning and Knowledge Extraction, 2025.
[205] Mauricio R. Delgado, V. Stenger, and J. Fiez. Motivation-dependent responses in the human caudate
nucleus. Cerebral Cortex, 2004.
[206] Xiang Deng, Yu Gu, Boyuan Zheng, Shijie Chen, Samuel Stevens, Boshi Wang, Huan Sun, and Yu Su.
Mind2web: Towards a generalist agent for the web. Neural Information Processing Systems, 2023.
[207] Yang Deng, Wenqiang Lei, Hongru Wang, and Tat seng Chua. Prompting and evaluating large lan-
guage models for proactive dialogues: Clarification, target-guided, and non-collaboration. Conference
on Empirical Methods in Natural Language Processing, 2023.
[208] Yang Deng, An Zhang, Yankai Lin, Xu Chen, Ji-Rong Wen, and Tat-Seng Chua. Large language
model powered agents in the web. The Web Conference, 2024.
[209] Yang Deng, Xuan Zhang, Wenxuan Zhang, Yifei Yuan, See-Kiong Ng, and Tat-Seng Chua. On the
multi-turn instruction following for conversational web agents. Annual Meeting of the Association for
Computational Linguistics, 2024.
[210] Brouillet Denis and Versace Rémy. The nature of the traces and the dynamics of memory. Psychology
and Behavioral Sciences, 2019.
[211] Mohammad Mahdi Derakhshani, Ivona Najdenkoska, Cees G. M. Snoek, M. Worring, and Yuki
Asano. Self-supervised open-ended classification with small visual language models, arXiv preprint
arXiv:2310.00500, 2023. URL https://arxiv.org/abs/2310.00500v2.
[212] Stefan Dernbach, Khushbu Agarwal, Alejandro Zuniga, Michael Henry, and Sutanay Choudhury.
Glam: Fine-tuning large language models for domain knowledge graph alignment via neighborhood
partitioning and generative subgraph encoding. AAAI Spring Symposia, 2024.
[213] Rushali Deshmukh, Rutuj Raut, Mayur Bhavsar, Sanika Gurav, and Y. Patil. Optimizing human-ai
interaction: Innovations in prompt engineering. 2025 3rd International Conference on Intelligent
Data Communication Technologies and Internet of Things (IDCIoT), 2025.
[214] Darshan Deshpande, Varun Gangal, Hersh Mehta, Jitin Krishnan, Anand Kannappan, and Rebecca
Qian. Trail: Trace reasoning and agentic issue localization, arXiv preprint arXiv:2505.08638, 2025.
URL https://arxiv.org/abs/2505.08638v3.
[215] Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. Bert: Pre-training of deep
bidirectional transformers for language understanding. North American Chapter of the Association
for Computational Linguistics, 2019.
[216] Dhruv Dhamani and Mary Lou Maher. The tyranny of possibilities in the design of task-oriented llm
systems: A scoping survey, arXiv preprint arXiv:2312.17601, 2023. URL https://arxiv.org/
abs/2312.17601v1.
[217] Frederick Dillon, Gregor Halvorsen, Simon Tattershall, Magnus Rowntree, and Gareth Vanderpool.
Contextual memory reweaving in large language models using layered latent state reconstruction,
arXiv preprint arXiv:2502.02046, 2025. URL https://arxiv.org/abs/2502.02046v2.
[218] Hanxing Ding, Shuchang Tao, Liang Pang, Zihao Wei, Jinyang Gao, Bolin Ding, Huawei Shen,
and Xueqi Chen. Toolcoder: A systematic code-empowered tool learning framework for large
language models, arXiv preprint arXiv:2502.11404, 2025. URL https://arxiv.org/abs/2502.
11404v2.
[219] Hongxin Ding, Yue Fang, Runchuan Zhu, Xinke Jiang, Jinyang Zhang, Yongxin Xu, Xu Chu, Junfeng
Zhao, and Yasha Wang. 3ds: Decomposed difficulty data selection’s case study on llm medical
domain adaptation. 2024.
[220] Jiayu Ding, Shuming Ma, Li Dong, Xingxing Zhang, Shaohan Huang, Wenhui Wang, and Furu Wei.
Longnet: Scaling transformers to 1, 000, 000, 000 tokens. arXiv preprint, 2023.
[221] Tianyu Ding, Tianyi Chen, Haidong Zhu, Jiachen Jiang, Yiqi Zhong, Jinxin Zhou, Guangzhi Wang,
Zhihui Zhu, Ilya Zharkov, and Luming Liang. The efficiency spectrum of large language models:
An algorithmic survey, arXiv preprint arXiv:2312.00678, 2023. URL https://arxiv.org/abs/
2312.00678v2.
[222] Yiran Ding, L. Zhang, Chengruidong Zhang, Yuanyuan Xu, Ning Shang, Jiahang Xu, Fan Yang,
and Mao Yang. Longrope: Extending llm context window beyond 2 million tokens. International
Conference on Machine Learning, 2024.
[223] Yiwen Ding, Zhiheng Xi, Wei He, Zhuoyuan Li, Yitao Zhai, Xiaowei Shi, Xunliang Cai, Tao Gui,
Qi Zhang, and Xuanjing Huang. Mitigating tail narrowing in llm self-improvement via socratic-guided
sampling. North American Chapter of the Association for Computational Linguistics, 2024.
[224] Christian Djeffal. Reflexive prompt engineering: A framework for responsible prompt engineering
and ai interaction design. Conference on Fairness, Accountability and Transparency, 2025.
[225] G Dong, Y Chen, X Li, J Jin, H Qian, and Y Zhu. . . . Tool-star: Empowering llm-brained multi-tool
reasoner via reinforcement learning. 2025. URL https://arxiv.org/abs/2505.16410.
[226] Guanting Dong, Jinxu Zhao, Tingfeng Hui, Daichi Guo, Wenlong Wan, Boqi Feng, Yueyan Qiu,
Zhuoma Gongque, Keqing He, Zechen Wang, and Weiran Xu. Revisit input perturbation problems
for llms: A unified robustness evaluation framework for noisy slot filling task. Natural Language
Processing and Chinese Computing, 2023.
[227] Guanting Dong, Yifei Chen, Xiaoxi Li, Jiajie Jin, Hongjin Qian, Yutao Zhu, Hangyu Mao, Guorui
Zhou, Zhicheng Dou, and Ji-Rong Wen. Tool-star: Empowering llm-brained multi-tool reasoner via
reinforcement learning. arXiv preprint, 2025.
[228] Kaiwen Dong. Large language model applied in multi-agent systema survey. Applied and Computa-
tional Engineering, 2024.
[229] Peijie Dong, Zhenheng Tang, Xiang-Hong Liu, Lujun Li, Xiaowen Chu, and Bo Li. Can compressed
llms truly act? an empirical evaluation of agentic capabilities in llm compression, arXiv preprint
arXiv:2505.19433, 2025. URL https://arxiv.org/abs/2505.19433v2.
[230] Vicky Dong, Hao Yu, and Yao Chen. Graph-augmented relation extraction model with llms-generated
support document, arXiv preprint arXiv:2410.23452, 2024. URL https://arxiv.org/abs/
2410.23452v1.
[231] Xiangjue Dong, Maria Teleki, and James Caverlee. A survey on llm inference-time self-improvement,
arXiv preprint arXiv:2412.14352, 2024. URL https://arxiv.org/abs/2412.14352v1.
[232] Yuxin Dong, Shuo Wang, Hongye Zheng, Jiajing Chen, Zhenhong Zhang, and Chihang Wang.
Advanced rag models with graph structures: Optimizing complex knowledge reasoning and text gen-
eration. 2024 5th International Symposium on Computer Engineering and Intelligent Communications
(ISCEIC), 2024.
[233] Zican Dong, Junyi Li, Xin Men, Wayne Xin Zhao, Bingbing Wang, Zhen Tian, Weipeng Chen, and
Ji-Rong Wen. Exploring context window of large language models via decomposed positional vectors.
Neural Information Processing Systems, 2024.
[234] Ehsan Doostmohammadi and Marco Kuhlmann. Studying the role of input-neighbor overlap in
retrieval-augmented language models training efficiency, arXiv preprint arXiv:2505.14309, 2025.
URL https://arxiv.org/abs/2505.14309v1.
[235] Mohammadreza Doostmohammadian, Alireza Aghasi, Mohammad Pirani, Ehsan Nekouei, H. Zarrabi,
Reza Keypour, Apostolos I. Rikos, and K. H. Johansson. Survey of distributed algorithms for resource
allocation over multi-agent systems, arXiv preprint arXiv:2401.15607, 2024. URL https://arxiv.
org/abs/2401.15607v1.
[236] A. Dorri, S. Kanhere, and R. Jurdak. Multi-agent systems: A survey. IEEE Access, 2018.
[237] Mauro Dragone. Component & service-based agent systems: Self-osgi. International Conference on
Agents and Artificial Intelligence, 2012.
[238] Alexandre Drouin, Maxime Gasse, Massimo Caccia, Issam H. Laradji, Manuel Del Verme, Tom
Marty, David Vazquez, Nicolas Chapados, and Alexandre Lacoste. WorkArena: How capable are web
agents at solving common knowledge work tasks? In Ruslan Salakhutdinov, Zico Kolter, Katherine
Heller, Adrian Weller, Nuria Oliver, Jonathan Scarlett, and Felix Berkenkamp, editors, Proceedings
of the 41st International Conference on Machine Learning, volume 235 of Proceedings of Machine
Learning Research, pages 11642–11662. PMLR, 21–27 Jul 2024. URL https://proceedings.
mlr.press/v235/drouin24a.html.
[239] Hung Du, Srikanth Thudumu, Rajesh Vasa, and K. Mouzakis. A survey on context-aware multi-agent
systems: Techniques, challenges and future directions, arXiv preprint arXiv:2402.01968, 2024. URL
https://arxiv.org/abs/2402.01968v2.
[240] Jingfei Du, Edouard Grave, Beliz Gunel, Vishrav Chaudhary, Onur Çelebi, Michael Auli, Ves Stoyanov,
and Alexis Conneau. Self-training improves pre-training for natural language understanding. North
American Chapter of the Association for Computational Linguistics, 2020.
[241] Jusen Du, Weigao Sun, Disen Lan, Jiaxi Hu, and Yu Cheng. Mom: Linear sequence modeling with
mixture-of-memories, arXiv preprint arXiv:2502.13685, 2025. URL https://arxiv.org/abs/
2502.13685v2.
[242] Mingxuan Du, Benfeng Xu, Chiwei Zhu, Xiaorui Wang, and Zhendong Mao. Deepresearch bench: A
comprehensive benchmark for deep research agents, arXiv preprint arXiv:2506.11763, 2025. URL
https://arxiv.org/abs/2506.11763v1.
[243] Shangheng Du, Jiabao Zhao, Jinxin Shi, Zhentao Xie, Xin Jiang, Yanhong Bai, and Liang He. A
survey on the optimization of large language model-based agents, arXiv preprint arXiv:2503.12434,
2025. URL https://arxiv.org/abs/2503.12434v1.
[244] Yiming Du, Hongru Wang, Zhengyi Zhao, Bin Liang, Baojun Wang, Wanjun Zhong, Zezhong Wang,
and Kam-Fai Wong. Perltqa: A personal long-term memory dataset for memory classification,
retrieval, and synthesis in question answering, arXiv preprint arXiv:2402.16288, 2024. URL https:
//arxiv.org/abs/2402.16288v1.
[245] Hanqi Duan, Yao Cheng, Jianxiang Yu, and Xiang Li. Can large language models act as ensembler
for multi-gnns?, arXiv preprint arXiv:2410.16822, 2024. URL https://arxiv.org/abs/2410.
16822v2.
[246] Peitong Duan, Chin yi Chen, Bjoern Hartmann, and Yang Li. Visual prompting with iterative
refinement for design critique generation, arXiv preprint arXiv:2412.16829, 2024. URL https:
//arxiv.org/abs/2412.16829v2.
[247] Brown Ebouky, A. Bartezzaghi, and Mattia Rigotti. Eliciting reasoning in language models with
cognitive tools, arXiv preprint arXiv:2506.12115, 2025. URL https://arxiv.org/abs/2506.
12115v1.
[248] Darren Edge, Ha Trinh, Newman Cheng, Joshua Bradley, Alex Chao, Apurva Mody, Steven Truitt,
Dasha Metropolitansky, Robert Osazuwa Ness, and Jonathan Larson. From local to global: A graph
rag approach to query-focused summarization. arXiv preprint arXiv:2404.16130, 2024.
[249] Candace Edwards. Hybrid context retrieval augmented generation pipeline: Llm-augmented
knowledge graphs and vector database for accreditation reporting assistance, arXiv preprint
arXiv:2405.15436, 2024. URL https://arxiv.org/abs/2405.15436v1.
[250] Abul Ehtesham, Aditi Singh, Gaurav Kumar Gupta, and Saket Kumar. A survey of agent interoperabil-
ity protocols: Model context protocol (mcp), agent communication protocol (acp), agent-to-agent
protocol (a2a), and agent network protocol (anp), arXiv preprint arXiv:2505.02279, 2025. URL
https://arxiv.org/abs/2505.02279v2.
[251] Will Epperson, Gagan Bansal, Victor Dibia, Adam Fourney, Jack Gerrits, Erkang Zhu, and Saleema
Amershi. Interactive debugging and steering of multi-agent ai systems. International Conference on
Human Factors in Computing Systems, 2025.
[252] Lutfi Eren Erdogan, Nicholas Lee, Siddharth Jha, Sehoon Kim, Ryan Tabrizi, Suhong Moon, Coleman
Hooper, G. Anumanchipalli, Kurt Keutzer, and A. Gholami. Tinyagent: Function calling at the edge.
Conference on Empirical Methods in Natural Language Processing, 2024.
[253] Oluwole Fagbohun, Rachel M. Harrison, and Anton Dereventsov. An empirical categorization of
prompting techniques for large language models: A practitioner’s guide. arXiv preprint, 2024.
[254] Kazem Faghih, Wenxiao Wang, Yize Cheng, Siddhant Bharti, Gaurang Sriramanan, S. Balasub-
ramanian, Parsa Hosseini, and S. Feizi. Gaming tool preferences in agentic llms, arXiv preprint
arXiv:2505.18135, 2025. URL https://arxiv.org/abs/2505.18135v1.
[255] Linxi (Jim) Fan, Guanzhi Wang, Yunfan Jiang, Ajay Mandlekar, Yuncong Yang, Haoyi Zhu, Andrew
Tang, De-An Huang, Yuke Zhu, and Anima Anandkumar. Minedojo: Building open-ended embodied
agents with internet-scale knowledge. Neural Information Processing Systems, 2022.
[256] Siqi Fan, Xiusheng Huang, Yiqun Yao, Xuezhi Fang, Kang Liu, Peng Han, Shuo Shang, Aixin Sun, and
Yequan Wang. If an llm were a character, would it know its own story? evaluating lifelong learning in
llms, arXiv preprint arXiv:2503.23514, 2025. URL https://arxiv.org/abs/2503.23514v1.
[257] Wenqi Fan, Yujuan Ding, Liang bo Ning, Shijie Wang, Hengyun Li, Dawei Yin, Tat-Seng Chua,
and Qing Li. A survey on rag meeting llms: Towards retrieval-augmented large language models.
Knowledge Discovery and Data Mining, 2024.
[258] Yue Fan, Xiaojian Ma, Rujie Wu, Yuntao Du, Jiaqi Li, Zhi Gao, and Qing Li. Videoagent: A memory-
augmented multimodal agent for video understanding, arXiv preprint arXiv:2403.11481, 2024. URL
https://arxiv.org/abs/2403.11481v2.
[259] Hongchao Fang and Pengtao Xie. An end-to-end contrastive self-supervised learning framework for
language understanding. Transactions of the Association for Computational Linguistics, 2022.
[260] Hongchao Fang, Sicheng Wang, Meng Zhou, Jiayuan Ding, and Pengtao Xie. Cert: Contrastive
self-supervised learning for language understanding, arXiv preprint arXiv:2005.12766, 2020. URL
https://arxiv.org/abs/2005.12766v2.
[261] Junfeng Fang, Zijun Yao, Ruipeng Wang, Haokai Ma, Xiang Wang, and Tat-Seng Chua. We
should identify and mitigate third-party safety risks in mcp-powered agent systems, arXiv preprint
arXiv:2506.13666, 2025. URL https://arxiv.org/abs/2506.13666v1.
[262] Siyuan Fang, Kaijing Ma, Tianyu Zheng, Xinrun Du, Ningxuan Lu, Ge Zhang, and Qingkun Tang.
Karpa: A training-free method of adapting knowledge graph as references for large language model’s
reasoning path aggregation. arXiv preprint, 2024.
[263] Wei-Wen Fang, Yang Zhang, Kaizhi Qian, James Glass, and Yada Zhu. Play2prompt: Zero-shot tool
instruction optimization for llm agents via tool play, arXiv preprint arXiv:2503.14432, 2025. URL
https://arxiv.org/abs/2503.14432v2.
[264] Yi Fang, Dongzhe Fan, D. Zha, and Qiaoyu Tan. Gaugllm: Improving graph contrastive learning for
text-attributed graphs with large language models. Knowledge Discovery and Data Mining, 2024.
[265] Yi Fang, Bowen Jin, Jiacheng Shen, Sirui Ding, Qiaoyu Tan, and Jiawei Han. Graphgpt-o: Synergistic
multimodal comprehension and generation on graphs. arXiv preprint, 2025.
[266] Bahare Fatemi, Jonathan J. Halcrow, and Bryan Perozzi. Talk like a graph: Encoding graphs for large
language models. International Conference on Learning Representations, 2023.
[267] George Fatouros, Georgios Makridis, George Kousiouris, John Soldatos, A. Tsadimas, and D. Kyriazis.
Towards conversational ai for human-machine collaborative mlops, arXiv preprint arXiv:2504.12477,
2025. URL https://arxiv.org/abs/2504.12477v1.
[268] M. Fauth, F. Wörgötter, and Christian Tetzlaff. Formation and maintenance of robust long-term
information storage in the presence of synaptic turnover. bioRxiv, 2015.
[269] Zahra Fayyaz, Aya Altamimi, Sen Cheng, and Laurenz Wiskott. A model of semantic completion in
generative episodic memory. Neural Computation, 2021.
[270] Xiang Fei, Xiawu Zheng, and Hao Feng. Mcp-zero: Proactive toolchain construction for llm agents
from scratch. arXiv preprint, 2025.
[271] Philip Feldman, James R. Foulds, and Shimei Pan. Ragged edges: The double-edged sword of
retrieval-augmented chatbots, arXiv preprint arXiv:2403.01193, 2024. URL https://arxiv.
org/abs/2403.01193v3.
[272] Aosong Feng, Rex Ying, and L. Tassiulas. Long sequence modeling with attention tensorization:
From sequence to tensor learning. Conference on Empirical Methods in Natural Language Processing,
2024.
[273] Erhu Feng, Wenbo Zhou, Zibin Liu, Le Chen, Yunpeng Dong, Cheng Zhang, Yisheng Zhao, Dong Du,
Zhi-Hua Zhou, Yubin Xia, and Haibo Chen. Get experience from practice: Llm agents with record &
replay, arXiv preprint arXiv:2505.17716, 2025. URL https://arxiv.org/abs/2505.17716v1.
[274] Jiazhan Feng, Shijue Huang, Xingwei Qu, Ge Zhang, Yujia Qin, Baoquan Zhong, Chengquan Jiang,
Jinxin Chi, and Wanjun Zhong. Retool: Reinforcement learning for strategic tool use in llms, arXiv
preprint arXiv:2504.11536, 2025. URL https://arxiv.org/abs/2504.11536v2.
[275] Kaituo Feng, Changsheng Li, Xiaolu Zhang, Jun Zhou, Ye Yuan, and Guoren Wang. Keypoint-based
progressive chain-of-thought distillation for llms. International Conference on Machine Learning,
2024.
[276] Leo Feng, Frederick Tung, Hossein Hajimirsadeghi, Y. Bengio, and M. O. Ahmed. Constant memory
attention block, arXiv preprint arXiv:2306.12599, 2023. URL https://arxiv.org/abs/2306.
12599v1.
[277] Yanlin Feng, Xinyue Chen, Bill Yuchen Lin, Peifeng Wang, Jun Yan, and Xiang Ren. Scalable multi-hop
relational reasoning for knowledge-aware question answering. Conference on Empirical Methods in
Natural Language Processing, 2020.
[278] Yifan Feng, Shiquan Liu, Xiangmin Han, Shaoyi Du, Zongze Wu, Han Hu, and Yue Gao. Hypergraph
foundation model, arXiv preprint arXiv:2503.01203v1, 2025. URL https://arxiv.org/abs/
2503.01203v1.
[279] Chrisantha Fernando, Dylan Banarse, H. Michalewski, Simon Osindero, and Tim Rocktäschel. Prompt-
breeder: Self-referential self-improvement via prompt evolution. International Conference on Machine
Learning, 2023.
[280] Tharindu Fernando, Simon Denman, A. Mcfadyen, S. Sridharan, and C. Fookes. Tree memory
networks for modelling long-term temporal dependencies. Neurocomputing, 2017.
[281] M. Ferrag, Norbert Tihanyi, and M. Debbah. From llm reasoning to autonomous ai agents: A
comprehensive review, arXiv preprint arXiv:2504.19678, 2025. URL https://arxiv.org/abs/
2504.19678v1.
[282] M. Ferrag, Norbert Tihanyi, and M. Debbah. Reasoning beyond limits: Advances and open prob-
lems for llms, arXiv preprint arXiv:2503.22732, 2025. URL https://arxiv.org/abs/2503.
22732v1.
[283] Christopher Fifty, Dennis Duan, Ronald G. Junkins, Ehsan Amid, Jurij Leskovec, Christopher R’e, and
Sebastian Thrun. Context-aware meta-learning. International Conference on Learning Representations,
2023.
[284] Tim Finin, Richard Fritzson, Donald P McKay, Robin McEntire, et al. Kqml-a language and protocol
for knowledge and information exchange. In 13th Int. Distributed Artificial Intelligence Workshop,
pages 93–103, 1994.
[285] Chelsea Finn, P. Abbeel, and S. Levine. Model-agnostic meta-learning for fast adaptation of deep
networks. International Conference on Machine Learning, 2017.
[286] Paolo Finotelli and Francis Eustache. Mathematical modeling of human memory. Frontiers in
Psychology, 2023.
[287] Ferdinando Fioretto, Enrico Pontelli, and W. Yeoh. Distributed constraint optimization problems and
applications: A survey. Journal of Artificial Intelligence Research, 2016.
[288] Meire Fortunato, Melissa Tan, Ryan Faulkner, S. Hansen, Adrià Puigdomènech Badia, Gavin Buttimore,
Charlie Deck, Joel Z. Leibo, and C. Blundell. Generalization of reinforcement learners with working
and episodic memory. Neural Information Processing Systems, 2019.
[289] Samy Foudil, Claire Pleche, and E. Macaluso. Memory for spatio-temporal contextual details during
the retrieval of naturalistic episodes. Scientific Reports, 2021.
[290] Zafeirios Fountas, Martin A Benfeghoul, Adnan Oomerjee, Fenia Christopoulou, Gerasimos Lam-
pouras, Haitham Bou-Ammar, and Jun Wang. Human-like episodic memory for infinite context llms,
arXiv preprint arXiv:2407.09450, 2024. URL https://arxiv.org/abs/2407.09450.
[291] Quentin Fournier, G. Caron, and D. Aloise. A practical survey on faster and lighter transformers.
ACM Computing Surveys, 2021.
[292] Luca Franceschi, P. Frasconi, Saverio Salzo, Riccardo Grazzi, and M. Pontil. Bilevel programming
for hyperparameter optimization and meta-learning. International Conference on Machine Learning,
2018.
[293] Eduard Frankford, Daniel Crazzolara, Clemens Sauerwein, Michael Vierhauser, and Ruth Breu.
Requirements for an online integrated development environment for automated programming
assessment systems. International Conference on Computer Supported Education, 2024.
[294] Chaoyou Fu, Yuhan Dai, Yondong Luo, Lei Li, Shuhuai Ren, Renrui Zhang, Zihan Wang, Chenyu
Zhou, Yunhang Shen, Mengdan Zhang, Peixian Chen, Yanwei Li, Shaohui Lin, Sirui Zhao, Ke Li,
Tong Xu, Xiawu Zheng, Enhong Chen, Rongrong Ji, and Xing Sun. Video-mme: The first-ever
comprehensive evaluation benchmark of multi-modal llms in video analysis. arXiv preprint, 2024.
[295] Honghao Fu, Yilang Shen, Yuxuan Liu, Jingzhong Li, and Xiang Zhang. Sgcn: a multi-order neigh-
borhood feature fusion landform classification method based on superpixel and graph convolutional
network. International Journal of Applied Earth Observation and Geoinformation, 122:103441, 2023.
[296] Honghao Fu, Yufei Wang, Wenhan Yang, Alex C Kot, and Bihan Wen. Dp-iqa: Utilizing diffusion
prior for blind image quality assessment in the wild. 2024.
[297] Honghao Fu, Hao Wang, Jing Jih Chin, and Zhiqi Shen. Brainvis: Exploring the bridge between brain
and visual signals via image reconstruction. In ICASSP 2025-2025 IEEE International Conference on
Acoustics, Speech and Signal Processing (ICASSP), pages 1–5. IEEE, 2025.
[298] Yuchuan Fu, Xiaohan Yuan, and Dongxia Wang. Ras-eval: A comprehensive benchmark for security
evaluation of llm agents in real-world environments. arXiv preprint, 2025.
[299] Zichuan Fu, Wentao Song, Yejing Wang, Xian Wu, Yefeng Zheng, Yingying Zhang, Derong Xu,
Xuetao Wei, Tong Xu, and Xiangyu Zhao. Sliding window attention training for efficient large
language models, arXiv preprint arXiv:2502.18845, 2025. URL https://arxiv.org/abs/2502.
18845v2.
[300] Stefano Fusi. Memory capacity of neural network models, arXiv preprint arXiv:2108.07839, 2021.
URL https://arxiv.org/abs/2108.07839v2.
[301] Tiantian Gan and Qiyao Sun. Rag-mcp: Mitigating prompt bloat in llm tool selection via retrieval-
augmented generation. arXiv preprint, 2025.
[302] Kanishk Gandhi, Gala Stojnic, B. Lake, and M. Dillon. Baby intuitions benchmark (bib): Discerning
the goals, preferences, and actions of others. Neural Information Processing Systems, 2021.
[303] Anish Ganguli, Prabal Deb, and Debleena Banerjee. Mark: Memory augmented refinement of knowl-
edge, arXiv preprint arXiv:2505.05177, 2025. URL https://arxiv.org/abs/2505.05177v1.
[304] Chen Gao, Xiaochong Lan, Nian Li, Yuan Yuan, Jingtao Ding, Zhilun Zhou, Fengli Xu, and Yong Li.
Large language models empowered agent-based modeling and simulation: A survey and perspectives.
Humanities and Social Sciences Communications, 2023.
[305] Chen Gao, Xiaochong Lan, Zhihong Lu, Jinzhu Mao, Jinghua Piao, Huandong Wang, Depeng Jin,
and Yong Li. S3 : Social-network simulation system with large language model-empowered agents,
arXiv preprint arXiv:2307.14984, 2025. URL https://arxiv.org/abs/2307.14984.
[306] Hang Gao and Yongfeng Zhang. Memory sharing for large language model based agents, arXiv
preprint arXiv:2404.09982, 2024. URL https://arxiv.org/abs/2404.09982v2.
[307] L Gao, A Madaan, S Zhou, and U Alon. . . . Pal: Program-aided language models. 2023. URL
https://proceedings.mlr.press/v202/gao23f.
[308] Luyu Gao, Xueguang Ma, Jimmy J. Lin, and Jamie Callan. Precise zero-shot dense retrieval without
relevance labels. Annual Meeting of the Association for Computational Linguistics, 2022.
[309] Luyu Gao, Aman Madaan, Shuyan Zhou, Uri Alon, Pengfei Liu, Yiming Yang, Jamie Callan, and
Graham Neubig. Pal: Program-aided language models. International Conference on Machine Learning,
2022.
[310] Shuzheng Gao, Xinjie Wen, Cuiyun Gao, Wenxuan Wang, and Michael R. Lyu. What makes good in-
context demonstrations for code intelligence tasks with llms? International Conference on Automated
Software Engineering, 2023.
[311] Tianyu Gao, Adam Fisch, and Danqi Chen. Making pre-trained language models better few-shot
learners. Annual Meeting of the Association for Computational Linguistics, 2021.
[312] Weiguo Gao. Mep: Multiple kernel learning enhancing relative positional encoding length extrapola-
tion, arXiv preprint arXiv:2403.17698, 2024. URL https://arxiv.org/abs/2403.17698v1.
[313] Xian Gao, Zongyun Zhang, Mingye Xie, Ting Liu, and Yuzhuo Fu. Graph of ai ideas: Leveraging
knowledge graphs and llms for ai research idea generation, arXiv preprint arXiv:2503.08549, 2025.
URL https://arxiv.org/abs/2503.08549v1.
[314] Xuanqi Gao, Siyi Xie, Juan Zhai, Shqing Ma, and Chao Shen. Mcp-radar: A multi-dimensional
benchmark for evaluating tool use capabilities in large language models. arXiv preprint, 2025.
[315] Yunfan Gao, Yun Xiong, Xinyu Gao, Kangxiang Jia, Jinliu Pan, Yuxi Bi, Yi Dai, Jiawei Sun, Qianyu Guo,
Meng Wang, and Haofen Wang. Retrieval-augmented generation for large language models: A survey,
arXiv preprint arXiv:2312.10997, 2023. URL https://arxiv.org/abs/2312.10997v5.
[316] Yunfan Gao, Yun Xiong, Meng Wang, and Haofen Wang. Modular rag: Transforming rag systems
into lego-like reconfigurable frameworks, arXiv preprint arXiv:2407.21059, 2024. URL https:
//arxiv.org/abs/2407.21059v1.
[317] Yunfan Gao, Yun Xiong, Yijie Zhong, Yuxi Bi, Ming Xue, and Haofen Wang. Synergizing rag and
reasoning: A systematic review, arXiv preprint arXiv:2504.15909, 2025. URL https://arxiv.
org/abs/2504.15909v2.
[318] Zhangyang Gao, Daize Dong, Cheng Tan, Jun Xia, Bozhen Hu, and Stan Z. Li. A graph is worth k
words: Euclideanizing graph using pure transformer. International Conference on Machine Learning,
2024.
[319] Itai Gat, Idan Schwartz, and A. Schwing. Perceptual score: What data modalities does your model
perceive? Neural Information Processing Systems, 2021.
[320] Itai Gat, Felix Kreuk, Tu Nguyen, Ann Lee, Jade Copet, Gabriel Synnaeve, Emmanuel Dupoux, and
Yossi Adi. Augmentation invariant discrete representation for generative spoken language modeling.
International Workshop on Spoken Language Translation, 2022.
[321] Tao Ge, Jing Hu, Xun Wang, Si-Qing Chen, and Furu Wei. In-context autoencoder for context
compression in a large language model. International Conference on Learning Representations, 2023.
[322] Yuyao Ge, Zhongguo Yang, Lizhe Chen, Yiming Wang, and Chengyang Li. Attack based on data: a
novel perspective to attack sensitive points directly. Cybersecurity, 6(1):43, 2023.
[323] Yuyao Ge, Shenghua Liu, Baolong Bi, Yiwei Wang, Lingrui Mei, Wenjie Feng, Lizhe Chen, and Xueqi
Cheng. Can graph descriptive order affect solving graph problems with llms? ACL 2025, 2024.
[324] Yuyao Ge, Shenghua Liu, Yiwei Wang, Lingrui Mei, Lizhe Chen, Baolong Bi, and Xueqi Cheng. Innate
reasoning is not enough: In-context learning enhances reasoning large language models with less
overthinking. 2025.
[325] Binzong Geng, Zhaoxin Huan, Xiaolu Zhang, Yong He, Liang Zhang, Fajie Yuan, Jun Zhou, and
Linjian Mo. Breaking the length barrier: Llm-enhanced ctr prediction in long textual user behaviors.
Annual International ACM SIGIR Conference on Research and Development in Information Retrieval,
2024.
[326] Hejia Geng, Boxun Xu, and Peng Li. Upar: A kantian-inspired prompting framework for enhancing
large language model capabilities, arXiv preprint arXiv:2310.01441, 2023. URL https://arxiv.
org/abs/2310.01441v2.
[327] Antonios Georgiou, M. Katkov, and M. Tsodyks. Retroactive interference model of forgetting. Journal
of Mathematical Neuroscience, 2021.
[328] S. Gershman, A. Schapiro, A. Hupbach, and K. Norman. Neural context reinstatement predicts
memory misattribution. Journal of Neuroscience, 2013.
[329] Mor Geva, R. Schuster, Jonathan Berant, and Omer Levy. Transformer feed-forward layers are
key-value memories. Conference on Empirical Methods in Natural Language Processing, 2020.
[330] Mor Geva, Avi Caciularu, Ke Wang, and Yoav Goldberg. Transformer feed-forward layers build
predictions by promoting concepts in the vocabulary space. Conference on Empirical Methods in
Natural Language Processing, 2022.
[331] Arda Gezdur and J. Bhattacharjya. Innovators and transformers: enhancing supply chain employee
training with an innovative application of a large language model. International Journal of Physical
Distribution & Logistics Management, 2025.
[332] Alireza Ghafarollahi and Markus J. Buehler. Protagents: protein discovery via large language model
multi-agent collaborations combining physics and machine learning. Digital Discovery, 2024.
[333] Abdellah Ghassel, Ian Robinson, Gabriel Tanase, Hal Cooper, Bryan Thompson, Zhen Han, V. Ioanni-
dis, Soji Adeshina, and H. Rangwala. Hierarchical lexical graph for enhanced multi-hop retrieval,
arXiv preprint arXiv:2506.08074, 2025. URL https://arxiv.org/abs/2506.08074v1.
[334] S. Ghetti and S. Bunge. Neural changes underlying the development of episodic memory during
middle childhood. Developmental Cognitive Neuroscience, 2012.
[335] D. Ghica. Function interface models for hardware compilation: Types, signatures, protocols, arXiv
preprint arXiv:0907.0749, 2009. URL https://arxiv.org/abs/0907.0749v1.
[336] Tyler Giallanza, Declan Campbell, and Jonathan D. Cohen. Toward the emergence of intelligent
control: Episodic generalization and optimization. Open Mind, 2024.
[337] In Gim, Seung seob Lee, and Lin Zhong. Asynchronous llm function calling, arXiv preprint
arXiv:2412.07017, 2024. URL https://arxiv.org/abs/2412.07017v1.
[338] Amelia Glaese, Nat McAleese, Maja Trębacz, John Aslanides, Vlad Firoiu, Timo Ewalds, Maribeth
Rauh, Laura Weidinger, Martin Chadwick, Phoebe Thacker, Lucy Campbell-Gillingham, Jonathan
Uesato, Po-Sen Huang, Ramona Comanescu, Fan Yang, Abigail See, Sumanth Dathathri, Rory
Greig, Charlie Chen, Doug Fritz, Jaume Sanchez Elias, Richard Green, Soňa Mokrá, Nicholas
Fernando, Boxi Wu, Rachel Foley, Susannah Young, Iason Gabriel, William Isaac, John Mellor, Demis
Hassabis, Koray Kavukcuoglu, Lisa Anne Hendricks, and Geoffrey Irving. Improving alignment of
dialogue agents via targeted human judgements, arXiv preprint arXiv:2209.14375, 2022. URL
https://arxiv.org/abs/2209.14375.
[339] D. Godden and A. Baddeley. Context-dependent memory in two natural environments: on land and
underwater. arXiv preprint, 1975.
[340] Arda Goknil, Femke B. Gelderblom, Simeon Tverdal, Shukun Tokas, and Hui Song. Privacy policy
analysis through prompt engineering for llms, arXiv preprint arXiv:2409.14879, 2024. URL https:
//arxiv.org/abs/2409.14879v1.
[341] Yaroslav Golubev, Zarina Kurbatova, E. Alomar, T. Bryksin, and Mohamed Wiem Mkaouer. One
thousand and one stories: a large-scale survey of software refactoring. ESEC/SIGSOFT FSE, 2021.
[342] Alan M Gordon, Jesse Rissman, Roozbeh Kiani, and Anthony D Wagner. Cortical reinstatement
mediates the relationship between content-specific encoding activity and subsequent recollection
decisions. Cerebral Cortex, 2014.
[343] E. Gordon and B. Logan. Managing goals and resources in dynamic environments. arXiv preprint,
2005.
[344] Z Gou, Z Shao, Y Gong, Y Shen, and Y Yang. . . . Critic: Large language models can self-correct with
tool-interactive critiquing. 2023. URL https://arxiv.org/abs/2305.11738.
[345] Zhibin Gou, Zhihong Shao, Yeyun Gong, Yelong Shen, Yujiu Yang, Minlie Huang, Nan Duan,
and Weizhu Chen. Tora: A tool-integrated reasoning agent for mathematical problem solving.
International Conference on Learning Representations, 2023.
[346] Alex Graves, Abdel rahman Mohamed, and Geoffrey E. Hinton. Speech recognition with deep
recurrent neural networks. IEEE International Conference on Acoustics, Speech, and Signal Processing,
2013.
[347] Ekaterina Grishina, Mikhail Gorbunov, and Maxim Rakhuba. Procrustesgpt: Compressing llms with
structured matrices and orthogonal transformations, arXiv preprint arXiv:2506.02818, 2025. URL
https://arxiv.org/abs/2506.02818v1.
[348] Sven Gronauer and K. Diepold. Multi-agent deep reinforcement learning: a survey. Artificial
Intelligence Review, 2021.
[349] C. Gros. Complex and adaptive dynamical systems, arXiv preprint arXiv:0807.4838, 2008. URL
https://arxiv.org/abs/0807.4838v3.
[350] Albert Gu, Karan Goel, and Christopher R’e. Efficiently modeling long sequences with structured
state spaces. International Conference on Learning Representations, 2021.
[351] Albert Gu, Ankit Gupta, Karan Goel, and Christopher Ré. On the parameterization and initialization
of diagonal state space models. Neural Information Processing Systems, 2022.
[352] Jian Gu, Chunyang Chen, and A. Aleti. Vocabulary-defined semantics: Latent space clustering for
improving in-context learning, arXiv preprint arXiv:2401.16184, 2024. URL https://arxiv.
org/abs/2401.16184v6.
[353] Yongli Gu, Xiang Yan, Hanlin Qin, Naveed Akhtar, Shuai Yuan, Honghao Fu, Shuowen Yang, and
Ajmal Mian. Hdtcnet: A hybrid-dimensional convolutional network for multivariate time series
classification. Pattern Recognition, page 111837, 2025.
[354] Zhuohan Gu, Jiayi Yao, Kuntai Du, and Junchen Jiang. Llmsteer: Improving long-context llm
inference by steering attention on reused contexts, arXiv preprint arXiv:2411.13009, 2024. URL
https://arxiv.org/abs/2411.13009v2.
[355] Shengyue Guan, Haoyi Xiong, Jindong Wang, Jiang Bian, Bin Zhu, and Jian guang Lou. Evaluating
llm-based agents for multi-turn conversations: A survey, arXiv preprint arXiv:2503.22458, 2025.
URL https://arxiv.org/abs/2503.22458v1.
[356] Zhong Guan, Hongke Zhao, Likang Wu, Ming He, and Jianpin Fan. Langtopo: Aligning language
descriptions of graphs with tokenized topological modeling, arXiv preprint arXiv:2406.13250, 2024.
URL https://arxiv.org/abs/2406.13250v1.
[357] Tom Gunter, Zirui Wang, Chong Wang, Ruoming Pang, Andy Narayanan, Aonan Zhang, Bowen
Zhang, Chen Chen, Chung-Cheng Chiu, David Qiu, Deepak Gopinath, Dian Ang Yap, Dong Yin,
Feng Nan, Floris Weers, Guoli Yin, Haoshuo Huang, Jianyu Wang, Jiarui Lu, John Peebles, Kewei
Ye, Mark Lee, Nan Du, Qibin Chen, Quentin Keunebroek, Sam Wiseman, Syd Evans, Tao Lei, Vivek
Rathod, Xiang Kong, Xianzhi Du, Yanghao Li, Yongqiang Wang, Yuan Gao, Zaid Ahmed, Zhaoyang
Xu, Zhiyun Lu, Al Rashid, Albin Madappally Jose, Alec Doane, Alfredo Bencomo, Allison Vanderby,
Andrew Hansen, Ankur Jain, A. Anupama, Areeba Kamal, Bugu Wu, Carolina Brum, Charlie Maalouf,
Chinguun Erdenebileg, Chris Dulhanty, Dominik Moritz, Doug Kang, Eduardo Jimenez, Evan Ladd,
Fang Shi, Felix Bai, Frank Chu, Fred Hohman, Hadas Kotek, Hannah Gillis Coleman, Jane Li, Jeffrey P.
Bigham, Jeffery Cao, Jeff Lai, Jessica Cheung, Jiulong Shan, Joe Zhou, John Li, Jun Qin, Karanjeet
Singh, Karla Vega, Kelvin Zou, Laura Heckman, Lauren Gardiner, Margit Bowler, Maria Cordell,
Meng Cao, Nicole Hay, Nilesh Shahdadpuri, Otto Godwin, Pranay Dighe, Pushyami Rachapudi,
Ramsey Tantawi, Roman Frigg, Sam Davarnia, Sanskruti Shah, Saptarshi Guha, Sasha Sirovica, Shen
Ma, Shuang Ma, Simon Wang, Sulgi Kim, Suma Jayaram, Vaishaal Shankar, Varsha Paidi, Vivek
Kumar, Xin Wang, Xin Zheng, Walker Cheng, Y. Shrager, Yang Ye, Yasu Tanaka, Yihao Guo, Yun
Meng, Zhaoping Luo, Ouyang Zhi, Alp Aygar, Alvin Wan, Andrew D. Walkingshaw, Tzu-Hsiang Lin,
Arsalan Farooq, Brent Ramerth, Colorado Reed, Chris Bartels, Chris Chaney, David Riazati, Eric Liang
Yang, Erin Feldman, Gabriel Hochstrasser, Guillaume Seguin, Irina Belousova, J. Pelemans, Karen
Yang, Keivan A. Vahid, Liangliang Cao, Mahyar Najibi, Marco Zuliani, Max Horton, Minsik Cho,
Nikhil Bhendawade, Patrick Dong, Piotr Maj, Pulkit Agrawal, Qi Shan, Qichen Fu, R. Poston, Sam
Xu, Shuangning Liu, Sushma Rao, Tashweena Heeramun, Thomas Merth, Uday Rayala, Victor Cui,
Vivek Rangarajan Sridhar, Wencong Zhang, Wenqi Zhang, Wentao Wu, Xingyu Zhou, Xinwen Liu,
Yang Zhao, Yin Xia, Zhile Ren, and Zhongzheng Ren. Apple intelligence foundation language models,
arXiv preprint arXiv:2407.21075, 2024. URL https://arxiv.org/abs/2407.21075v1.
[358] Jiayan Guo, Lun Du, and Hengyu Liu. Gpt4graph: Can large language models understand graph
structured data ? an empirical evaluation and benchmarking, arXiv preprint arXiv:2305.15066,
2023. URL https://arxiv.org/abs/2305.15066v2.
[359] Jing Guo, Nan Li, Jianchuan Qi, Hang Yang, Ruiqiao Li, Yuzhen Feng, Si Zhang, and Ming Xu.
Empowering working memory for large language model agents, arXiv preprint arXiv:2312.17259,
2024. URL https://arxiv.org/abs/2312.17259.
[360] Taicheng Guo, Xiuying Chen, Yaqi Wang, Ruidi Chang, Shichao Pei, N. Chawla, Olaf Wiest, and
Xiangliang Zhang. Large language model based multi-agents: A survey of progress and challenges.
International Joint Conference on Artificial Intelligence, 2024.
[361] Xiaojun Guo, Ang Li, Yifei Wang, Stefanie Jegelka, and Yisen Wang. G1: Teaching llms to reason on
graphs with reinforcement learning. arXiv preprint arXiv:2505.18499, 2025.
[362] Yuan Guo, Tingjia Miao, Zheng Wu, Pengzhou Cheng, Ming Zhou, and Zhuosheng Zhang. Atomic-
to-compositional generalization for mobile agents with a new benchmark and scheduling system,
arXiv preprint arXiv:2506.08972, 2025. URL https://arxiv.org/abs/2506.08972v1.
[363] Zhicheng Guo, Sijie Cheng, Hao Wang, Shihao Liang, Yujia Qin, Peng Li, Zhiyuan Liu, Maosong Sun,
and Yang Liu. Stabletoolbench: Towards stable large-scale benchmarking on tool learning of large
language models. Annual Meeting of the Association for Computational Linguistics, 2024.
[364] Zirui Guo, Lianghao Xia, Yanhua Yu, Tu Ao, and Chao Huang. Lightrag: Simple and fast retrieval-
augmented generation, arXiv preprint arXiv:2410.05779, 2024. URL https://arxiv.org/abs/
2410.05779v3.
[365] Sharut Gupta, Chenyu Wang, Yifei Wang, T. Jaakkola, and Stefanie Jegelka. In-context symmetries:
Self-supervised learning through contextual world models. Neural Information Processing Systems,
2024.
[366] Tanmay Gupta, Luca Weihs, and Aniruddha Kembhavi. Codenav: Beyond tool-use to using real-world
codebases with llm agents, arXiv preprint arXiv:2406.12276, 2024. URL https://arxiv.org/
abs/2406.12276v1.
[367] I Gur, H Furuta, A Huang, M Safdari, and Y Matsuo. . . . A real-world webagent with planning, long
context understanding, and program synthesis. 2023. URL https://arxiv.org/abs/2307.
12856.
[368] Izzeddin Gur, Hiroki Furuta, Austin Huang, Mustafa Safdari, Yutaka Matsuo, D. Eck, and Aleksandra
Faust. A real-world webagent with planning, long context understanding, and program synthesis.
International Conference on Learning Representations, 2023.
[369] Ashwin Kumar Gururajan, Enrique Lopez-Cuena, Jordi Bayarri-Planas, Adrián Tormos, Daniel Hinjos,
Pablo Bernabeu Perez, Anna Arias-Duart, Pablo A. Martin-Torres, Lucia Urcelay-Ganzabal, Marta
Gonzalez-Mallo, S. Álvarez Napagao, Eduard Ayguad’e-Parra, and Ulises Cortés Dario Garcia-Gasulla.
Aloe: A family of fine-tuned open healthcare llms, arXiv preprint arXiv:2405.01886, 2024. URL
https://arxiv.org/abs/2405.01886v1.
[370] Bernal Jimenez Gutierrez, Yiheng Shu, Yu Gu, Michihiro Yasunaga, and Yu Su. Hipporag: Neuro-
biologically inspired long-term memory for large language models. Neural Information Processing
Systems, 2024.
[371] Kelvin Guu, Kenton Lee, Zora Tung, Panupong Pasupat, and Ming-Wei Chang. Realm: Retrieval-
augmented language model pre-training. International Conference on Machine Learning, 2020.
[372] K. Gödel, B. Meltzer, and R. Schlegel. On formally undecidable propositions of principia mathematica
and related systems. arXiv preprint, 1966.
[373] Idan Habler, Ken Huang, Vineeth Sai Narajala, and Prashant Kulkarni. Building a secure agentic
ai application leveraging a2a protocol, arXiv preprint arXiv:2504.16902, 2025. URL https://
arxiv.org/abs/2504.16902v2.
[374] John Halloran. Mcp safety training: Learning to refuse falsely benign mcp exploits using improved
preference alignment, arXiv preprint arXiv:2505.23634, 2025. URL https://arxiv.org/abs/
2505.23634v1.
[375] Tae Jun Ham, Yejin Lee, Seong Hoon Seo, Soo-Uck Kim, Hyunji Choi, Sungjun Jung, and Jae W.
Lee. Elsa: Hardware-software co-design for efficient, lightweight self-attention mechanism in neural
networks. International Symposium on Computer Architecture, 2021.
[376] Feijiang Han, Licheng Guo, Hengtao Cui, and Zhiyuan Lyu. Question tokens deserve more attention:
Enhancing large language models without training through step-by-step reading and question
attention recalibration, arXiv preprint arXiv:2504.09402, 2025. URL https://arxiv.org/abs/
2504.09402v1.
[377] Han Han, Tong Zhu, Xiang Zhang, Mengsong Wu, Hao Xiong, and Wenliang Chen. Nestools: A
dataset for evaluating nested tool learning abilities of large language models. International Conference
on Computational Linguistics, 2024.
[378] Haoyu Han, Yu Wang, Harry Shomer, Kai Guo, Jiayuan Ding, Yongjia Lei, Mahantesh Halappanavar,
Ryan A. Rossi, Subhabrata Mukherjee, Xianfeng Tang, Qi He, Zhigang Hua, Bo Long, Tong Zhao,
Neil Shah, Amin Javari, Yinglong Xia, and Jiliang Tang. Retrieval-augmented generation with
graphs (graphrag), arXiv preprint arXiv:2501.00309, 2025. URL https://arxiv.org/abs/
2501.00309.
[379] Tingxu Han, Zhenting Wang, Chunrong Fang, Shiyun Zhao, Shiqing Ma, and Zhenyu Chen. Token-
budget-aware llm reasoning, arXiv preprint arXiv:2412.18547, 2024. URL https://arxiv.org/
abs/2412.18547v5.
[380] Yuanning Han, Ziyi Qiu, Jiale Cheng, and Ray Lc. When teams embrace ai: Human collaboration
strategies in generative prompting in a creative design task. International Conference on Human
Factors in Computing Systems, 2024.
[381] R. Hankache, Kingsley Nketia Acheampong, Liang Song, Marek Brynda, Raad Khraishi, and Greig A.
Cowan. Evaluating the sensitivity of llms to prior context, arXiv preprint arXiv:2506.00069, 2025.
URL https://arxiv.org/abs/2506.00069v1.
[382] S Hao, T Liu, Z Wang, and Z Hu. Toolkengpt: Augmenting frozen language models with massive
tools via tool embeddings. 2023. URL https://proceedings.neurips.cc/paper_files/
paper/2023/hash/8fd1a81c882cd45f64958da6284f4a3f-Abstract-Conference.
html.
[383] Mohanakrishnan Hariharan. Semantic mastery: Enhancing llms with advanced natural language
understanding, arXiv preprint arXiv:2504.00409, 2025. URL https://arxiv.org/abs/2504.
00409v1.
[384] Mareike Hartmann and Alexander Koller. A survey on complex tasks for goal-directed interac-
tive agents, arXiv preprint arXiv:2409.18538, 2024. URL https://arxiv.org/abs/2409.
18538v1.
[385] A. Hassani, A. Medvedev, P. D. Haghighi, Sea Ling, A. Zaslavsky, and P. Jayaraman. Context definition
and query language: Conceptual specification, implementation, and evaluation. Italian National
Conference on Sensors, 2019.
[386] Kostas Hatalis, Despina Christou, Joshua Myers, Steven Jones, Keith Lambert, Adam Amos-Binks,
Zohreh Dannenhauer, and Dustin Dannenhauer. Memory matters: The need to improve long-term
memory in llm-agents. Proceedings of the AAAI Symposium Series, 2024.
[387] Kostas Hatalis, Despina Christou, and Vyshnavi Kondapalli. Review of case-based reasoning for llm
agents: Theoretical foundations, architectural components, and cognitive integration, arXiv preprint
arXiv:2504.06943, 2025. URL https://arxiv.org/abs/2504.06943v2.
[388] Jacky He, Guiran Liu, Binrong Zhu, Hanlu Zhang, Hongye Zheng, and Xiaokai Wang. Context-guided
dynamic retrieval for improving generation quality in rag models, arXiv preprint arXiv:2504.19436,
2025. URL https://arxiv.org/abs/2504.19436v1.
[389] Jianben He, Xingbo Wang, Shiyi Liu, Guande Wu, Claudio Silva, and Huamin Qu. Poem: Interactive
prompt optimization for enhancing multimodal reasoning of large language models. IEEE Pacific
Visualization Symposium, 2024.
[390] Junqing He, Liang Zhu, Rui Wang, Xi Wang, Gholamreza Haffari, and Jiaxing Zhang. Madial-bench:
Towards real-world evaluation of memory-augmented dialogue generation. North American Chapter
of the Association for Computational Linguistics, 2024.
[391] Shawn He, Surangika Ranathunga, Stephen Cranefield, and B. Savarimuthu. Norm violation
detection in multi-agent systems using large language models: A pilot study. COINE, 2024.
[392] Shengtao He. Achieving tool calling functionality in llms using only prompt engineering with-
out fine-tuning, arXiv preprint arXiv:2407.04997, 2024. URL https://arxiv.org/abs/2407.
04997v1.
[393] Wenchong He, Liqian Peng, Zhe Jiang, and Alex Go. You only fine-tune once: Many-shot in-
context fine-tuning for large language model, arXiv preprint arXiv:2506.11103, 2025. URL https:
//arxiv.org/abs/2506.11103v1.
[394] Xu He, Di Wu, Yan Zhai, and Kun Sun. Sentinelagent: Graph-based anomaly detection in multi-
agent systems, arXiv preprint arXiv:2505.24201, 2025. URL https://arxiv.org/abs/2505.
24201v1.
[395] Yang He, Xiao Ding, Bibo Cai, Yufei Zhang, Kai Xiong, Zhouhao Sun, Bing Qin, and Ting Liu. Self-
route: Automatic mode switching via capability estimation for efficient reasoning. arXiv preprint,
2025.
[396] Yu He, Yingxi Li, Colin White, and Ellen Vitercik. Dsr-bench: Evaluating the structural reasoning
abilities of llms via data structures. arXiv preprint, 2025.
[397] Zexue He, Leonid Karlinsky, Donghyun Kim, Julian McAuley, Dmitry Krotov, and Rogério Feris.
Camelot: Towards large language models with training-free consolidated associative memory, arXiv
preprint arXiv:2402.13449, 2024. URL https://arxiv.org/abs/2402.13449v1.
[398] James B. Heald, M. Lengyel, and D. Wolpert. Contextual inference in learning and memory. Trends
in Cognitive Sciences, 2022.
[399] Shekoofeh Hedayati, Ryan E. O’Donnell, and Brad Wyble. A model of working memory for latent
representations. Nature Human Behaviour, 2021.
[400] Tooraj Helmi. Modeling response consistency in multi-agent llm systems: A comparative analysis
of shared and separate context approaches, arXiv preprint arXiv:2504.07303, 2025. URL https:
//arxiv.org/abs/2504.07303v1.
[401] Arshia Hemmat, Kianoosh Vadaei, Mohammad Hassan Heydari, and Afsaneh Fatemi. Leveraging
retrieval-augmented generation for persian university knowledge retrieval. Conference on Information
and Knowledge Technology, 2024.
[402] M. Herrera, Marco Pérez-Hernández, A. Kumar Parlikad, and J. Izquierdo. Multi-agent systems and
complex networks: Review and applications in systems engineering. Processes, 2020.
[403] Nora A. Herweg, A. Sharan, M. Sperling, A. Brandt, A. Schulze-Bonhage, and M. Kahana. Reactivated
spatial context guides episodic recall. Journal of Neuroscience, 2018.
[404] Thomas F. Heston and Charya Khun. Prompt engineering in medical education. International Medical
Education, 2023.
[405] Dollaya Hirunyasiri, Danielle R. Thomas, Jionghao Lin, K. Koedinger, and Vincent Aleven. Com-
parative analysis of gpt-4 and human graders in evaluating human tutors giving praise to students.
Human-AI Math Tutoring@AIED, 2023.
[406] Thomas Hoang. Gnn: Graph neural network and large language model for data discovery, arXiv
preprint arXiv:2408.13609, 2024. URL https://arxiv.org/abs/2408.13609v2.
[407] W. Hoek and M. Wooldridge. Towards a logic of rational agency. Logic Journal of the IGPL, 2003.
[408] Aidan Hogan, E. Blomqvist, Michael Cochez, C. d’Amato, Gerard de Melo, C. Gutierrez, J. E. L. Gayo,
S. Kirrane, S. Neumaier, A. Polleres, Roberto Navigli, A. Ngomo, S. M. Rashid, Anisa Rula, Lukas
Schmelzeisen, Juan Sequeda, Steffen Staab, and Antoine Zimmermann. Knowledge graphs. ACM
Computing Surveys, 2020.
[409] Nithin Holla, Pushkar Mishra, H. Yannakoudakis, and Ekaterina Shutova. Meta-learning with sparse
experience replay for lifelong language learning, arXiv preprint arXiv:2009.04891, 2020. URL
https://arxiv.org/abs/2009.04891v2.
[410] Chuanyang Hong and Qingyun He. Enhancing memory retrieval in generative agents through
llm-trained cross attention networks. Frontiers in Psychology, 2025.
[411] M. Hong, Sean M. Polyn, and Lisa K. Fazio. Examining the episodic context account: does retrieval
practice enhance memory for context? Cognitive Research, 2019.
[412] Sirui Hong, Xiawu Zheng, Jonathan P. Chen, Yuheng Cheng, Ceyao Zhang, Zili Wang, Steven
Ka Shing Yau, Z. Lin, Liyang Zhou, Chenyu Ran, Lingfeng Xiao, and Chenglin Wu. Metagpt: Meta
programming for multi-agent collaborative framework. arXiv preprint, 2023.
[413] Sirui Hong, Mingchen Zhuge, Jiaqi Chen, Xiawu Zheng, Yuheng Cheng, Ceyao Zhang, Jinlin Wang,
Zili Wang, Steven Ka Shing Yau, Zijuan Lin, Liyang Zhou, Chenyu Ran, Lingfeng Xiao, Chenglin Wu,
and Jürgen Schmidhuber. Metagpt: Meta programming for a multi-agent collaborative framework,
arXiv preprint arXiv:2308.00352, 2024. URL https://arxiv.org/abs/2308.00352.
[414] Wenyi Hong, Weihan Wang, Qingsong Lv, Jiazheng Xu, Wenmeng Yu, Junhui Ji, Yan Wang, Zihan
Wang, Yuxiao Dong, Ming Ding, and Jie Tang. Cogagent: A visual language model for gui agents.
Computer Vision and Pattern Recognition, 2023.
[415] Xiangyu Hong, Che Jiang, Biqing Qi, Fandong Meng, Mo Yu, Bowen Zhou, and Jie Zhou. On the
token distance modeling ability of higher rope attention dimension. Conference on Empirical Methods
in Natural Language Processing, 2024.
[416] Yubin Hong, Chaofan Li, Jingyi Zhang, and Yingxia Shao. Fg-rag: Enhancing query-focused summa-
rization with context-aware fine-grained graph rag. arXiv preprint, 2025.
[417] Thanapapas Horsuwan, Piyawat Lertvittayakumjorn, Kasidis Kanwatchara, B. Kijsirikul, and P. Va-
teekul. Meta lifelong-learning with selective and task-aware adaptation. IEEE Access, 2024.
[418] A. N. Hoskin, A. Bornstein, K. Norman, and J. Cohen. Refresh my memory: Episodic memory rein-
statements intrude on working memory maintenance. Cognitive, Affective, & Behavioral Neuroscience,
2017.
[419] Timothy M. Hospedales, Antreas Antoniou, P. Micaelli, and A. Storkey. Meta-learning in neural
networks: A survey. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020.
[420] Peyman Hosseini, Ignacio Castro, Iacopo Ghinassi, and Matthew Purver. Efficient solutions for an
intriguing failure of llms: Long context window does not mean llms can analyze long sequences
flawlessly. International Conference on Computational Linguistics, 2024.
[421] Haowen Hou, Fei Ma, Binwen Bai, Xinxin Zhu, and F. Yu. Enhancing and accelerating large language
models via instruction-aware contextual compression, arXiv preprint arXiv:2408.15491, 2024. URL
https://arxiv.org/abs/2408.15491v1.
[422] Wenjun Hou, Yi Cheng, Kaishuai Xu, Yan Hu, Wenjie Li, and Jiangming Liu. Memory-augmented
multimodal llms for surgical vqa via self-contained inquiry, arXiv preprint arXiv:2411.10937v1, 2024.
URL https://arxiv.org/abs/2411.10937v1.
[423] Xinyi Hou, Yanjie Zhao, Shenao Wang, and Haoyu Wang. Model context protocol (mcp): Landscape,
security threats, and future research directions, arXiv preprint arXiv:2503.23278, 2025. URL
https://arxiv.org/abs/2503.23278v2.
[424] Zejiang Hou, Julian Salazar, and George Polovets. Meta-learning the difference: Preparing large
language models for efficient adaptation. Transactions of the Association for Computational Linguistics,
2022.
[425] N. Houlsby, A. Giurgiu, Stanislaw Jastrzebski, Bruna Morrone, Quentin de Laroussilhe, Andrea
Gesmundo, Mona Attariyan, and S. Gelly. Parameter-efficient transfer learning for nlp. International
Conference on Machine Learning, 2019.
[426] Marc W Howard and M. Kahana. A distributed representation of temporal context. arXiv preprint,
2002.
[427] Chenxu Hu, Jie Fu, Chenzhuang Du, Simian Luo, J. Zhao, and Hang Zhao. Chatdb: Augmenting
llms with databases as their symbolic memory, arXiv preprint arXiv:2306.03901, 2023. URL https:
//arxiv.org/abs/2306.03901v2.
[428] J. E. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, and Weizhu
Chen. Lora: Low-rank adaptation of large language models. International Conference on Learning
Representations, 2021.
[429] Junhao Hu, Wenrui Huang, Weidong Wang, Zhenwen Li, Tiancheng Hu, Zhixia Liu, XuSheng Chen,
Tao Xie, and Yizhou Shan. Raas: Reasoning-aware attention sparsity for efficient llm reasoning,
arXiv preprint arXiv:2502.11147, 2025. URL https://arxiv.org/abs/2502.11147v2.
[430] Junwei Hu, Weicheng Zheng, Yan Liu, and Yihan Liu. Optimizing token consumption in llms: A
nano surge approach for code reasoning efficiency, arXiv preprint arXiv:2504.15989, 2025. URL
https://arxiv.org/abs/2504.15989v2.
[431] Junyan Hu, Hanlin Niu, J. Carrasco, B. Lennox, and F. Arvin. Voronoi-based multi-robot autonomous
exploration in unknown environments via deep reinforcement learning. IEEE Transactions on
Vehicular Technology, 2020.
[432] Linmei Hu, Zeyi Liu, Ziwang Zhao, Lei Hou, Liqiang Nie, and Juanzi Li. A survey of knowledge
enhanced pre-trained language models. IEEE Transactions on Knowledge and Data Engineering, 2022.
[433] Mengkang Hu, Tianxing Chen, Qiguang Chen, Yao Mu, Wenqi Shao, and Ping Luo. Hiagent:
Hierarchical working memory management for solving long-horizon agent tasks with large language
model, arXiv preprint arXiv:2408.09559, 2024. URL https://arxiv.org/abs/2408.09559.
[434] Mengkang Hu, Yao Mu, Xinmiao Yu, Mingyu Ding, Shiguang Wu, Wenqi Shao, Qiguang Chen, Bin
Wang, Yu Qiao, and Ping Luo. Tree-planner: Efficient close-loop task planning with large language
models, arXiv preprint arXiv:2310.08582, 2024. URL https://arxiv.org/abs/2310.08582.
[435] Mengkang Hu, Yuhang Zhou, Wendong Fan, Yuzhou Nie, Bowei Xia, Tao Sun, Ziyu Ye, Zhaoxuan
Jin, Yingru Li, Qiguang Chen, Zeyu Zhang, Yifeng Wang, Qianshuo Ye, Bernard Ghanem, Ping Luo,
and Guohao Li. Owl: Optimized workforce learning for general multi-agent assistance in real-world
task automation, arXiv preprint arXiv:2505.23885, 2025. URL https://arxiv.org/abs/2505.
23885.
[436] Nathan J. Hu, E. Mitchell, Christopher D. Manning, and Chelsea Finn. Meta-learning online
adaptation of language models. Conference on Empirical Methods in Natural Language Processing,
2023.
[437] Shengxiang Hu, Guobing Zou, Song Yang, Yanglan Gan, Bofeng Zhang, and Yixin Chen. Large
language model meets graph neural network in knowledge distillation. AAAI Conference on Artificial
Intelligence, 2024.
[438] Siyuan Hu, Mingyu Ouyang, Difei Gao, and Mike Zheng Shou. The dawn of gui agent: A preliminary
case study with claude 3.5 computer use. arXiv preprint, 2024.
[439] Ting Hu, Christoph Meinel, and Haojin Yang. Scaled prompt-tuning for few-shot natural lan-
guage generation, arXiv preprint arXiv:2309.06759, 2023. URL https://arxiv.org/abs/
2309.06759v1.
[440] Xiang Hu, Hongyu Fu, Jinge Wang, Yifeng Wang, Zhikun Li, Renjun Xu, Yu Lu, Yaochu Jin, Lili Pan,
and Zhenzhong Lan. Nova: An iterative planning and search approach to enhance novelty and
diversity of llm generated ideas, arXiv preprint arXiv:2410.14255, 2024. URL https://arxiv.
org/abs/2410.14255v2.
[441] Yuntong Hu, Zhengwu Zhang, and Liang Zhao. Beyond text: A deep dive into large language
models’ ability on understanding graph data, arXiv preprint arXiv:2310.04944, 2023. URL https:
//arxiv.org/abs/2310.04944v1.
[442] Yilun Hua and Yoav Artzi. Talk less, interact better: Evaluating in-context conversational adaptation
in multimodal llms, arXiv preprint arXiv:2408.01417v1, 2024. URL https://arxiv.org/abs/
2408.01417v1.
[443] Brandon Huang, Chancharik Mitra, Assaf Arbelle, Leonid Karlinsky, Trevor Darrell, and Roei Herzig.
Multimodal task vectors enable many-shot multimodal in-context learning. Neural Information
Processing Systems, 2024.
[444] Chengkai Huang, Hongtao Huang, Tong Yu, Kaige Xie, Junda Wu, Shuai Zhang, Julian J. McAuley,
Dietmar Jannach, and Lina Yao. A survey of foundation model-powered recommender systems:
From feature-based, generative to agentic paradigms, arXiv preprint arXiv:2504.16420, 2025. URL
https://arxiv.org/abs/2504.16420v1.
[445] Chengkai Huang, Junda Wu, Yu Xia, Zixu Yu, Ruhan Wang, Tong Yu, Ruiyi Zhang, Ryan A. Rossi,
B. Kveton, Dongruo Zhou, Julian J. McAuley, and Lina Yao. Towards agentic recommender systems
in the era of multimodal large language models, arXiv preprint arXiv:2503.16734, 2025. URL
https://arxiv.org/abs/2503.16734v1.
[446] Chengrui Huang, Shen Gao, Zhengliang Shi, Dongsheng Wang, and Shuo Shang. Ttpa: Token-
level tool-use preference alignment training framework with fine-grained evaluation, arXiv preprint
arXiv:2505.20016, 2025. URL https://arxiv.org/abs/2505.20016v1.
[447] Chensen Huang, Guibo Zhu, Xuepeng Wang, Yifei Luo, Guojing Ge, Haoran Chen, Dong Yi, and
Jinqiao Wang. Recurrent context compression: Efficiently expanding the context window of llm,
arXiv preprint arXiv:2406.06110, 2024. URL https://arxiv.org/abs/2406.06110v1.
[448] Jing Huang, X. Ruan, Naigong Yu, Qingwu Fan, Jiaming Li, and Jianxian Cai. A cognitive model
based on neuromodulated plasticity. Computational Intelligence and Neuroscience, 2016.
[449] Ken Huang, Akram Sheriff, Vineeth Sai Narajala, and Idan Habler. Agent capability negotiation and
binding protocol (acnbp), arXiv preprint arXiv:2506.13590, 2025. URL https://arxiv.org/
abs/2506.13590v1.
[450] Le Huang, Hengzhi Lan, Zijun Sun, Chuan Shi, and Ting Bai. Emotional rag: Enhancing role-playing
agents through emotional retrieval. 2024 IEEE International Conference on Knowledge Graph (ICKG),
2024.
[451] Lisheng Huang, Yichen Liu, Jinhao Jiang, Rongxiang Zhang, Jiahao Yan, Junyi Li, and Wayne Xin
Zhao. Manusearch: Democratizing deep search in large language models with a transparent and
open multi-agent framework, arXiv preprint arXiv:2505.18105, 2025. URL https://arxiv.org/
abs/2505.18105v1.
[452] Shiting Huang, Zhen Fang, Zehui Chen, Siyu Yuan, Junjie Ye, Yu Zeng, Lin Chen, Qi Mao, and
Feng Zhao. Critictool: Evaluating self-critique capabilities of large language models in tool-calling
error scenarios, arXiv preprint arXiv:2506.13977, 2025. URL https://arxiv.org/abs/2506.
13977v1.
[453] Sirui Huang, Yanggan Gu, Xuming Hu, Zhonghao Li, Qing Li, and Guandong Xu. Reasoning factual
knowledge in structured data with large language models, arXiv preprint arXiv:2408.12188, 2024.
URL https://arxiv.org/abs/2408.12188v1.
[454] Sirui Huang, Hanqian Li, Yanggan Gu, Xuming Hu, Qing Li, and Guandong Xu. Hyperg: Hypergraph-
enhanced llms for structured knowledge, arXiv preprint arXiv:2502.18125, 2025. URL https:
//arxiv.org/abs/2502.18125v1.
[455] Wenlong Huang, P. Abbeel, Deepak Pathak, and Igor Mordatch. Language models as zero-shot
planners: Extracting actionable knowledge for embodied agents. International Conference on Machine
Learning, 2022.
[456] Xu Huang, Jianxun Lian, Yuxuan Lei, Jing Yao, Defu Lian, and Xing Xie. Recommender ai agent:
Integrating large language models for interactive recommendations, arXiv preprint arXiv:2308.16505,
2024. URL https://arxiv.org/abs/2308.16505.
[457] Xu Huang, Weiwen Liu, Xiaolong Chen, Xingmei Wang, Hao Wang, Defu Lian, Yasheng Wang,
Ruiming Tang, and Enhong Chen. Understanding the planning of llm agents: A survey, arXiv preprint
arXiv:2402.02716, 2024. URL https://arxiv.org/abs/2402.02716v1.
[458] Y Huang, J Shi, Y Li, C Fan, S Wu, and Q Zhang. . . . Metatool benchmark for large language models:
Deciding whether to use tools and which to use. 2023. URL https://arxiv.org/abs/2310.
03128.
[459] Yunpeng Huang, Jingwei Xu, Zixu Jiang, Junyu Lai, Zenan Li, Yuan Yao, Taolue Chen, Lijuan
Yang, Zhou Xin, and Xiaoxing Ma. Advancing transformer architecture in long-context large
language models: A comprehensive survey, arXiv preprint arXiv:2311.12351, 2023. URL https:
//arxiv.org/abs/2311.12351v2.
[460] Zeyi Huang, Yuyang Ji, Anirudh Sundara Rajan, Zefan Cai, Wen Xiao, Junjie Hu, and Yong Jae Lee.
Visualtoolagent (vista): A reinforcement learning framework for visual tool selection, arXiv preprint
arXiv:2505.20289, 2025. URL https://arxiv.org/abs/2505.20289v1.
[461] Ziheng Huang, S. Gutierrez, Hemanth Kamana, and S. Macneil. Memory sandbox: Transparent
and interactive memory management for conversational agents. ACM Symposium on User Interface
Software and Technology, 2023.
[462] Ziheng Huang, Sebastian Gutierrez, Hemanth Kamana, and Stephen MacNeil. Memory sand-
box: Transparent and interactive memory management for conversational agents, arXiv preprint
arXiv:2308.01542, 2023. URL https://arxiv.org/abs/2308.01542.
[463] Alexis Huet, Zied Ben-Houidi, and Dario Rossi. Episodic memories generation and evaluation
benchmark for large language models. International Conference on Learning Representations, 2025.
[464] Dom Huh and Prasant Mohapatra. Multi-agent reinforcement learning: A comprehensive survey,
arXiv preprint arXiv:2312.10256, 2023. URL https://arxiv.org/abs/2312.10256v2.
[465] Eunjeong Hwang, Yichao Zhou, James Bradley Wendt, Beliz Gunel, Nguyen Vo, Jing Xie, and Sandeep
Tata. Enhancing incremental summarization with structured representations. Conference on Empirical
Methods in Natural Language Processing, 2024.
[466] Thorsten Händler. Balancing autonomy and alignment: A multi-dimensional taxonomy for au-
tonomous llm-powered multi-agent architectures. arXiv preprint, 2023.
[467] Michael Iannelli, Sneha Kuchipudi, and Vera Dvorak. Sla management in reconfigurable multi-agent
rag: A systems approach to question answering, arXiv preprint arXiv:2412.06832, 2024. URL
https://arxiv.org/abs/2412.06832v2.
[468] IBM. What is agent communication protocol (acp)? https://www.ibm.com/think/topics/
agent-communication-protocol, 2025. [Online; accessed 17-July-2025].
[469] T Inaba, H Kiyomaru, F Cheng, and S Kurohashi. Multitool-cot: Gpt-3 can use multiple external
tools with chain of thought prompting. 2023. URL https://arxiv.org/abs/2305.16896.
[470] G. Indiveri and Shih-Chii Liu. Memory and information processing in neuromorphic systems.
Proceedings of the IEEE, 2015.
[471] V. Ioannidis, Xiang Song, Da Zheng, Houyu Zhang, Jun Ma, Yi Xu, Belinda Zeng, Trishul M. Chilimbi,
and G. Karypis. Efficient and effective training of language and graph neural network models, arXiv
preprint arXiv:2206.10781, 2022. URL https://arxiv.org/abs/2206.10781v1.
[472] Yoichi Ishibashi, Taro Yano, and M. Oyamada. Can large language models invent algorithms to im-
prove themselves?: Algorithm discovery for recursive self-improvement through reinforcement learn-
ing, arXiv preprint arXiv:2410.15639, 2024. URL https://arxiv.org/abs/2410.15639v5.
[473] Shadi Iskander, Nachshon Cohen, Zohar S. Karnin, Ori Shapira, and Sofia Tolmach. Quality matters:
Evaluating synthetic data for tool-using llms. Conference on Empirical Methods in Natural Language
Processing, 2024.
[474] Z. Ismail and N. Sariff. A survey and analysis of cooperative multi-agent robot systems: Challenges
and directions. Applications of Mobile Robots, 2018.
[475] Yusuf Izmirlioglu, Loc Pham, Tran Cao Son, and Enrico Pontelli. A survey of multi-agent systems for
smartgrids. Energies, 2024.
[476] Jace.AI. Jace.ai web agent, 2024. URL https://www.jace.ai/. Accessed: 2025-07-14.
[477] Arthur Jacot, Franck Gabriel, and Clément Hongler. Neural tangent kernel: Convergence and
generalization in neural networks. Neural Information Processing Systems, 2018.
[478] Tejas Jade and Alex Yartsev. Chatgpt for automated grading of short answer questions in mechani-
cal ventilation, arXiv preprint arXiv:2505.04645, 2025. URL https://arxiv.org/abs/2505.
04645v1.
[479] A. Jafarpour, L. Fuentemilla, A. Horner, W. Penny, and E. Duzel. Replay of very early encoding
representations during recollection. Journal of Neuroscience, 2014.
[480] A. Jaiswal, Nurendra Choudhary, Ravinarayana Adkathimar, M. P. Alagappan, G. Hiranandani,
Ying Ding, Zhangyang Wang, E-Wen Huang, and Karthik Subbian. All against some: Efficient
integration of large language models for message passing in graph neural networks, arXiv preprint
arXiv:2407.14996, 2024. URL https://arxiv.org/abs/2407.14996v1.
[481] H. Jaleel, Jane J. Stephan, and Sinan Naji. Multi-agent systems: A review study. Ibn AL- Haitham
Journal For Pure and Applied Sciences, 2020.
[482] Lawrence Jang, Yinheng Li, Charles Ding, Justin Lin, Paul Pu Liang, Dan Zhao, Rogerio Bonatti, and
K. Koishida. Videowebarena: Evaluating long context multimodal agents with video understanding
web tasks. International Conference on Learning Representations, 2024.
[483] R. Janik. Aspects of human memory and large language models, arXiv preprint arXiv:2311.03839,
2023. URL https://arxiv.org/abs/2311.03839v3.
[484] Shumaila Javaid, Hamza Fahim, Bin He, and Nasir Saeed. Large language models for uavs: Current
state and pathways to the future. IEEE Open Journal of Vehicular Technology, 2024.
[485] T. S. Jayram, Younes Bouhadjar, Ryan L. McAvoy, Tomasz Kornuta, Alexis Asseman, K. Rocki, and
A. Ozcan. Learning to remember, forget and ignore using attention control in memory, arXiv preprint
arXiv:1809.11087, 2018. URL https://arxiv.org/abs/1809.11087v1.
[486] Cheonsu Jeong. A study on the mcp x a2a framework for enhancing interoperability of llm-based
autonomous agents, arXiv preprint arXiv:2506.01804, 2025. URL https://arxiv.org/abs/
2506.01804v2.
[487] Gang Ji and J. Bilmes. Multi-speaker language modeling. North American Chapter of the Association
for Computational Linguistics, 2004.
[488] Ke Ji, Junying Chen, Anningzhe Gao, Wenya Xie, Xiang Wan, and Benyou Wang. Llms could
autonomously learn without external supervision, arXiv preprint arXiv:2406.00606, 2024. URL
https://arxiv.org/abs/2406.00606v2.
[489] Shaoxiong Ji, Shirui Pan, E. Cambria, P. Marttinen, and Philip S. Yu. A survey on knowledge graphs:
Representation, acquisition, and applications. IEEE Transactions on Neural Networks and Learning
Systems, 2020.
[490] Bowen Jiang, Runchuan Zhu, Jiang Wu, Zinco Jiang, Yifan He, Junyuan Gao, Jia Yu, Rui Min, Yinfan
Wang, Haote Yang, et al. Evaluating large language model with knowledge oriented language
specific simple question answering. 2025.
[491] Caigao Jiang, Siqiao Xue, James Zhang, Lingyue Liu, Zhibo Zhu, and Hongyan Hao. Learning large-
scale universal user representation with sparse mixture of experts, arXiv preprint arXiv:2207.04648,
2022. URL https://arxiv.org/abs/2207.04648v1.
[492] Feibo Jiang, Li Dong, Yubo Peng, Kezhi Wang, Kun Yang, Cunhua Pan, D. Niyato, and O. Dobre. Large
language model enhanced multi-agent systems for 6g communications. IEEE wireless communications,
2023.
[493] J Jiang, K Zhou, WX Zhao, Y Song, and C Zhu. . . . Kg-agent: An efficient autonomous agent
framework for complex reasoning over knowledge graph. 2024. URL https://arxiv.org/abs/
2402.11163.
[494] Jinhao Jiang, Kun Zhou, Wayne Xin Zhao, and Ji rong Wen. Unikgqa: Unified retrieval and reasoning
for solving multi-hop question answering over knowledge graph. International Conference on Learning
Representations, 2022.
[495] Jinhao Jiang, Kun Zhou, Zican Dong, Keming Ye, Wayne Xin Zhao, and Ji rong Wen. Structgpt: A
general framework for large language model to reason over structured data. Conference on Empirical
Methods in Natural Language Processing, 2023.
[496] Song Jiang, Zahra Shakeri, Aaron Chan, Maziar Sanjabi, Hamed Firooz, Yinglong Xia, Bugra Akyildiz,
Yizhou Sun, Jinchao Li, Qifan Wang, and Asli Celikyilmaz. Resprompt: Residual connection prompting
advances multi-step reasoning in large language models. North American Chapter of the Association
for Computational Linguistics, 2023.
[497] Zhengbao Jiang, Frank F. Xu, Luyu Gao, Zhiqing Sun, Qian Liu, Jane Dwivedi-Yu, Yiming Yang,
Jamie Callan, and Graham Neubig. Active retrieval augmented generation. Conference on Empirical
Methods in Natural Language Processing, 2023.
[498] Zhonglin Jiang, Qian Tang, and Zequn Wang. Generative reliability-based design optimization using
in-context learning capabilities of large language models, arXiv preprint arXiv:2503.22401, 2025.
URL https://arxiv.org/abs/2503.22401v1.
[499] Zhuoxuan Jiang, Haoyuan Peng, Shanshan Feng, Fan Li, and Dongsheng Li. Llms can find mathemat-
ical reasoning mistakes by pedagogical chain-of-thought. International Joint Conference on Artificial
Intelligence, 2024.
[500] Carlos E. Jimenez, John Yang, Alexander Wettig, Shunyu Yao, Kexin Pei, Ofir Press, and Karthik
Narasimhan. Swe-bench: Can language models resolve real-world github issues?, arXiv preprint
arXiv:2310.06770, 2024. URL https://arxiv.org/abs/2310.06770.
[501] B Jin, C Xie, J Zhang, KK Roy, Y Zhang, and Z Li. . . . Graph chain-of-thought: Augmenting large
language models by reasoning on graphs. 2024. URL https://arxiv.org/abs/2404.07103.
[502] Bowen Jin, Yu Zhang, Qi Zhu, and Jiawei Han. Heterformer: Transformer-based deep node
representation learning on heterogeneous text-rich networks. Knowledge Discovery and Data Mining,
2022.
[503] Feihu Jin, Jiajun Zhang, and Chengqing Zong. Parameter-efficient tuning for large language model
without calculating its gradients. Conference on Empirical Methods in Natural Language Processing,
2023.
[504] Haolin Jin, Linghan Huang, Haipeng Cai, Jun Yan, Bo Li, and Huaming Chen. From llms to llm-
based agents for software engineering: A survey of current, challenges and future, arXiv preprint
arXiv:2408.02479, 2024. URL https://arxiv.org/abs/2408.02479v2.
[505] Hongye Jin, Xiaotian Han, Jingfeng Yang, Zhimeng Jiang, Zirui Liu, Chia yuan Chang, Huiyuan
Chen, and Xia Hu. Llm maybe longlm: Self-extend llm context window without tuning. International
Conference on Machine Learning, 2024.
[506] Jiajie Jin, Yutao Zhu, Xinyu Yang, Chenghao Zhang, and Zhicheng Dou. Flashrag: A modular toolkit
for efficient retrieval-augmented generation research. The Web Conference, 2024.
[507] Qiao Jin, Yifan Yang, Qingyu Chen, and Zhiyong Lu. Genegpt: Augmenting large language models
with domain tools for improved access to biomedical information. Bioinform., 2023.
[508] Tian Jin, W. Yazar, Zifei Xu, Sayeh Sharify, and Xin Wang. Self-selected attention span for accelerating
large language model inference, arXiv preprint arXiv:2404.09336, 2024. URL https://arxiv.
org/abs/2404.09336v1.
[509] Weiqiang Jin, Hongyang Du, Biao Zhao, Xingwu Tian, Bohang Shi, and Guang Yang. A compre-
hensive survey on multi-agent cooperative decision-making: Scenarios, approaches, challenges and
perspectives. arXiv preprint, 2025.
[510] Yang Jin, Kun Xu, Kun Xu, Liwei Chen, Chao Liao, Jianchao Tan, Quzhe Huang, Bin Chen, Chenyi
Lei, An Liu, Chengru Song, Xiaoqiang Lei, Di Zhang, Wenwu Ou, Kun Gai, and Yadong Mu. Unified
language-vision pretraining in llm with dynamic discrete visual tokenization. International Conference
on Learning Representations, 2023.
[511] Yiqiao Jin, Minje Choi, Gaurav Verma, Jindong Wang, and Srijan Kumar. Mm-soc: Benchmarking
multimodal large language models in social media platforms. Annual Meeting of the Association for
Computational Linguistics, 2024.
[512] Yiyang Jin, Kunzhao Xu, Hang Li, Xueting Han, Yanmin Zhou, Cheng Li, and Jing Bai. Reveal:
Self-evolving code agents via iterative generation-verification, arXiv preprint arXiv:2506.11442,
2025. URL https://arxiv.org/abs/2506.11442v1.
[513] Jeff Johnson, Matthijs Douze, and H. Jégou. Billion-scale similarity search with gpus. IEEE Transac-
tions on Big Data, 2017.
[514] Jeff A. Johnson and Daniel H. Bullock. Fragility in ais using artificial neural networks. Communications
of the ACM, 2023.
[515] Zhao Kaiya, Michelangelo Naim, J. Kondic, Manuel Cortes, Jiaxin Ge, Shuying Luo, Guangyu Robert
Yang, and Andrew Ahn. Lyfe agents: Generative agents for low-cost real-time social interactions,
arXiv preprint arXiv:2310.02172, 2023. URL https://arxiv.org/abs/2310.02172v1.
[516] Kurmanbek Kaiyrbekov, Nic Dobbins, and Sean D. Mooney. Automated survey collection with llm-
based conversational agents, arXiv preprint arXiv:2504.02891, 2025. URL https://arxiv.org/
abs/2504.02891v1.
[517] A. Kakas, P. Mancarella, F. Sadri, Kostas Stathis, and Francesca Toni. Computational logic foundations
of kgp agents. Journal of Artificial Intelligence Research, 2008.
[518] Vikas Kamra, Lakshya Gupta, Dhruv Arora, and Ashwin Kumar Yadav. Enhancing document retrieval
using ai and graph-based rag techniques. 2024 5th International Conference on Communication,
Computing & Industry 6.0 (C2I6), 2024.
[519] Eser Kandogan, Nikita Bhutani, Dan Zhang, Rafael Li Chen, Sairam Gurajada, and Estevam R.
Hruschka. Orchestrating agents and data for enterprise: A blueprint architecture for compound ai,
arXiv preprint arXiv:2504.08148, 2025. URL https://arxiv.org/abs/2504.08148v1.
[520] Haoyu Kang, Yuzhou Zhu, Yukun Zhong, Ke Wang Central South University, Dalian University
of Technology, Nanjing University, and Xidian University. Sakr: Enhancing retrieval-augmented
generation via streaming algorithm and k-means clustering. arXiv preprint, 2024.
[521] Jiazheng Kang, Mingming Ji, Zhe Zhao, and Ting Bai. Memory os of ai agent, arXiv preprint
arXiv:2506.06326, 2025. URL https://arxiv.org/abs/2506.06326v1.
[522] Jikun Kang, Wenqi Wu, Filippos Christianos, Alex J. Chan, Fraser Greenlee, George Thomas, Marvin
Purtorab, and Andy Toulis. Lm2: Large memory models, arXiv preprint arXiv:2502.06049, 2025.
URL https://arxiv.org/abs/2502.06049v1.
[523] Sungmin Kang, Gabin An, and S. Yoo. A quantitative and qualitative evaluation of llm-based
explainable fault localization. Proc. ACM Softw. Eng., 2023.
[524] Guy Kaplan, Matanel Oren, Yuval Reif, and Roy Schwartz. From tokens to words: On the inner
lexicon of llms. International Conference on Learning Representations, 2024.
[525] Vladimir Karpukhin, Barlas Oğuz, Sewon Min, Patrick Lewis, Ledell Yu Wu, Sergey Edunov, Danqi
Chen, and Wen tau Yih. Dense passage retrieval for open-domain question answering. Conference on
Empirical Methods in Natural Language Processing, 2020.
[526] Zdeněk Kasner and Ondrej Dusek. Beyond traditional benchmarks: Analyzing behaviors of open llms
on data-to-text generation. Annual Meeting of the Association for Computational Linguistics, 2024.
[527] Kiran Kate, Tejaswini Pedapati, Kinjal Basu, Yara Rizk, Vijil Chenthamarakshan, Subhajit Chaudhury,
Mayank Agarwal, and Ibrahim Abdelaziz. Longfunceval: Measuring the effectiveness of long context
models for function calling, arXiv preprint arXiv:2505.10570, 2025. URL https://arxiv.org/
abs/2505.10570v1.
[528] Angelos Katharopoulos, Apoorv Vyas, Nikolaos Pappas, and Franccois Fleuret. Transformers are
rnns: Fast autoregressive transformers with linear attention. International Conference on Machine
Learning, 2020.
[529] Richard Katrix, Quentin Carroway, Rowan Hawkesbury, and Matthias Heathfield. Context-aware
semantic recomposition mechanism for large language models, arXiv preprint arXiv:2501.17386,
2025. URL https://arxiv.org/abs/2501.17386v2.
[530] Amirhossein Kazemnejad, Inkit Padhi, K. Ramamurthy, Payel Das, and Siva Reddy. The impact of
positional encoding on length generalization in transformers. Neural Information Processing Systems,
2023.
[531] T. Kelley, R. Thomson, and Jonathan Milton. Standard model of mind: Episodic memory. Biologically
Inspired Cognitive Architectures, 2018.
[532] Daan Kepel and Konstantina Valogianni. Autonomous prompt engineering in large language models,
arXiv preprint arXiv:2407.11000, 2024. URL https://arxiv.org/abs/2407.11000v1.
[533] R. Kesner. Neurobiological foundations of an attribute model of memory. arXiv preprint, 2013.
[534] A. Khan, Md Toufique Hasan, Kai-Kristian Kemell, Jussi Rasku, and Pekka Abrahamsson. Developing
retrieval augmented generation (rag) based llm systems from pdfs: An experience report, arXiv
preprint arXiv:2410.15944, 2024. URL https://arxiv.org/abs/2410.15944v1.
[535] Muhammad Tayyab Khan, Lequn Chen, Ye Han Ng, Wenhe Feng, Nicholas Yew Jin Tan, and Seung Ki
Moon. Leveraging vision-language models for manufacturing feature recognition in cad designs,
arXiv preprint arXiv:2411.02810, 2024. URL https://arxiv.org/abs/2411.02810v1.
[536] Urvashi Khandelwal, Omer Levy, Dan Jurafsky, Luke Zettlemoyer, and M. Lewis. Generalization
through memorization: Nearest neighbor language models. International Conference on Learning
Representations, 2019.
[537] Elahe Khatibi, Ziyu Wang, and Amir M. Rahmani. Cdf-rag: Causal dynamic feedback for adaptive
retrieval-augmented generation. arXiv preprint, 2025.
[538] Prannay Khosla, Piotr Teterwak, Chen Wang, Aaron Sarna, Yonglong Tian, Phillip Isola, Aaron
Maschinot, Ce Liu, and Dilip Krishnan. Supervised contrastive learning. Neural Information Processing
Systems, 2020.
[539] Tushar Khot, H. Trivedi, Matthew Finlayson, Yao Fu, Kyle Richardson, Peter Clark, and Ashish
Sabharwal. Decomposed prompting: A modular approach for solving complex tasks. International
Conference on Learning Representations, 2022.
[540] Sambhav Khurana, Xiner Li, Shurui Gui, and Shuiwang Ji. A hierarchical language model for
interpretable graph reasoning, arXiv preprint arXiv:2410.22372, 2024. URL https://arxiv.
org/abs/2410.22372v1.
[541] Daehee Kim, Deokhyung Kang, Sangwon Ryu, and Gary Geunbae Lee. Ontology-free general-domain
knowledge graph-to-text generation dataset synthesis using large language model, arXiv preprint
arXiv:2409.07088, 2024. URL https://arxiv.org/abs/2409.07088v1.
[542] Geunwoo Kim, P. Baldi, and S. McAleer. Language models can solve computer tasks. Neural
Information Processing Systems, 2023.
[543] Jaeyeon Kim, Injune Hwang, and Kyogu Lee. Learning semantic information from raw audio signal
using both contextual and phonetic representations. IEEE International Conference on Acoustics,
Speech, and Signal Processing, 2024.
[544] Jang-Hyun Kim, Junyoung Yeom, Sangdoo Yun, and Hyun Oh Song. Compressed context memory
for online language model interaction. International Conference on Learning Representations, 2023.
[545] Jiho Kim, Yeonsu Kwon, Yohan Jo, and Edward Choi. Kg-gpt: A general framework for reasoning
on knowledge graphs using large language models. Conference on Empirical Methods in Natural
Language Processing, 2023.
[546] Jiin Kim, Byeongjun Shin, Jinha Chung, and Minsoo Rhu. The cost of dynamic reasoning: De-
mystifying ai agents and test-time scaling from an ai infrastructure perspective, arXiv preprint
arXiv:2506.04301, 2025. URL https://arxiv.org/abs/2506.04301v1.
[547] Sehoon Kim, Suhong Moon, Ryan Tabrizi, Nicholas Lee, Michael W. Mahoney, Kurt Keutzer, and
A. Gholami. An llm compiler for parallel function calling. International Conference on Machine
Learning, 2023.
[548] Taewoon Kim, Michael Cochez, Vincent Francois-Lavet, Mark Neerincx, and Piek Vossen. A machine
with short-term, episodic, and semantic memory systems. Proceedings of the AAAI Conference on
Artificial Intelligence, 37(1):48–56, 2023. ISSN 2159-5399. doi: 10.1609/aaai.v37i1.25075. URL
http://dx.doi.org/10.1609/aaai.v37i1.25075.
[549] Lukas Kirchdorfer, Robert Blümel, T. Kampik, Han van der Aa, and Heiner Stuckenschmidt. Dis-
covering multi-agent systems for resource-centric business process simulation. Process Science,
2025.
[550] J. Kirkpatrick, Razvan Pascanu, Neil C. Rabinowitz, J. Veness, Guillaume Desjardins, Andrei A.
Rusu, Kieran Milan, John Quan, Tiago Ramalho, A. Grabska-Barwinska, D. Hassabis, C. Clopath,
D. Kumaran, and R. Hadsell. Overcoming catastrophic forgetting in neural networks. Proceedings of
the National Academy of Sciences of the United States of America, 2016.
[551] Louis Kirsch, James Harrison, Jascha Narain Sohl-Dickstein, and Luke Metz. General-purpose
in-context learning by meta-learning transformers, arXiv preprint arXiv:2212.04458, 2022. URL
https://arxiv.org/abs/2212.04458v2.
[552] Yuval Kirstain, Patrick Lewis, Sebastian Riedel, and Omer Levy. A few more examples may be worth
billions of parameters. Conference on Empirical Methods in Natural Language Processing, 2021.
[553] Andrew Kiruluta, Preethi Raju, and Priscilla Burity. Breaking quadratic barriers: A non-attention
llm for ultra-long context horizons, arXiv preprint arXiv:2506.01963, 2025. URL https://arxiv.
org/abs/2506.01963v1.
[554] Nikita Kitaev, Lukasz Kaiser, and Anselm Levskaya. Reformer: The efficient transformer. International
Conference on Learning Representations, 2020.
[555] Vincent Koc, Jacques Verre, Douglas Blank, and Abigail Morgan. Mind the metrics: Patterns for
telemetry-aware in-ide ai application development using the model context protocol (mcp), arXiv
preprint arXiv:2506.11019, 2025. URL https://arxiv.org/abs/2506.11019v1.
[556] Tomás Kociský, Jonathan Schwarz, Phil Blunsom, Chris Dyer, Karl Moritz Hermann, Gábor Melis,
and Edward Grefenstette. The narrativeqa reading comprehension challenge. Transactions of the
Association for Computational Linguistics, 2017.
[557] Jing Yu Koh, R. Salakhutdinov, and Daniel Fried. Grounding language models to images for
multimodal inputs and outputs. International Conference on Machine Learning, 2023.
[558] Jing Yu Koh, Robert Lo, Lawrence Jang, Vikram Duvvur, Ming Chong Lim, Po-Yu Huang, Gra-
ham Neubig, Shuyan Zhou, Ruslan Salakhutdinov, and Daniel Fried. Visualwebarena: Evaluating
multimodal agents on realistic visual web tasks, arXiv preprint arXiv:2401.13649, 2024. URL
https://arxiv.org/abs/2401.13649v2.
[559] Takeshi Kojima, S. Gu, Machel Reid, Yutaka Matsuo, and Yusuke Iwasawa. Large language models
are zero-shot reasoners. Neural Information Processing Systems, 2022.
[560] Yilun Kong, Jingqing Ruan, Yihong Chen, Bin Zhang, Tianpeng Bao, Shiwei Shi, Guoqing Du, Xiaoru
Hu, Hangyu Mao, Ziyue Li, Xingyu Zeng, and Rui Zhao. Tptu-v2: Boosting task planning and tool
usage of large language model-based agents in real-world systems, arXiv preprint arXiv:2311.11315,
2023. URL https://arxiv.org/abs/2311.11315.
[561] P. Korzyński, G. Mazurek, Pamela Krzypkowska, and Artur Kurasiński. Artificial intelligence prompt
engineering as a new digital competence: Analysis of generative ai technologies such as chatgpt.
Entrepreneurial Business and Economics Review, 2023.
[562] Oliver Kramer. Cognitive prompts using guilford’s structure of intellect model. arXiv preprint, 2025.
[563] Oliver Kramer. Conceptual metaphor theory as a prompting paradigm for large language models,
arXiv preprint arXiv:2502.01901, 2025. URL https://arxiv.org/abs/2502.01901v1.
[564] Oliver Kramer and Jill Baumann. Unlocking structured thinking in language models with cognitive
prompting. ESANN 2025 proceesdings, 2024.
[565] K. Kravari and Nick Bassiliades. A survey of agent platforms. Journal of Artificial Societies and Social
Simulation, 2015.
[566] Prashant Krishnan, Zilong Wang, Yangkun Wang, and Jingbo Shang. Towards few-shot entity
recognition in document images: A graph neural network approach robust to image manipulation.
International Conference on Language Resources and Evaluation, 2023.
[567] W. Kruijne, S. Bohté, P. Roelfsema, and C. Olivers. Flexible working memory through selective gating
and attentional tagging. bioRxiv, 2019.
[568] L. Krupp, Daniel Geissler, P. Lukowicz, and Jakob Karolus. Towards sustainable web agents: A plea
for transparency and dedicated metrics for energy consumption, arXiv preprint arXiv:2502.17903,
2025. URL https://arxiv.org/abs/2502.17903v1.
[569] M. Kuhail, Jose Berengueres, Fatma Taher, Sana Z. Khan, and Ansah Siddiqui. Designing a haptic
boot for space with prompt engineering: Process, insights, and implications. IEEE Access, 2024.
[570] Amandeep Kumar, Muzammal Naseer, Sanath Narayan, R. Anwer, Salman H. Khan, and
Hisham Cholakkal. Multi-modal generation via cross-modal in-context learning, arXiv preprint
arXiv:2405.18304v1, 2024. URL https://arxiv.org/abs/2405.18304v1.
[571] Rajeev Kumar, Harishankar Kumar, and Kumari Shalini. Detecting and mitigating bias in llms through
knowledge graph-augmented training. 2025 International Conference on Artificial Intelligence and
Data Engineering (AIDE), 2025.
[572] Taeyoon Kwon, Dongwook Choi, Sunghwan Kim, Hyojun Kim, Seungjun Moon, Beong woo Kwak,
Kuan-Hao Huang, and Jinyoung Yeo. Embodied agents meet personalization: Exploring memory
utilization for personalized assistance, arXiv preprint arXiv:2505.16348, 2025. URL https://
arxiv.org/abs/2505.16348v1.
[573] Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, Ying Sheng, Lianmin Zheng, Cody Hao Yu, Joseph E.
Gonzalez, Haotong Zhang, and Ion Stoica. Efficient memory management for large language model
serving with pagedattention. Symposium on Operating Systems Principles, 2023.
[574] T. Lai, Quan Hung Tran, Trung Bui, and D. Kihara. A gated self-attention memory network for
answer selection. Conference on Empirical Methods in Natural Language Processing, 2019.
[575] Divya Lamba. The role of prompt engineering in improving language understanding and generation.
International Journal For Multidisciplinary Research, 2024.
[576] Xiaochong Lan, Jie Feng, Jia Lei, Xinlei Shi, and Yong Li. Benchmarking and advancing large
language models for local life services, arXiv preprint arXiv:2506.02720, 2025. URL https:
//arxiv.org/abs/2506.02720v1.
[577] LangChain Team. Memory in langgraph. https://langchain-ai.github.io/langgraph/
concepts/memory/, 2025. Accessed: 2025-07-17.
[578] Samuel T. Langlois, Oghenetekevwe Akoroda, Estefany Carrillo, J. Herrmann, S. Azarm, Huan Xu,
and Michael W. Otte. Metareasoning structures, problems, and modes for multiagent systems: A
survey. IEEE Access, 2020.
[579] B. Lattimer, Varun Gangal, Ryan McDonald, and Yi Yang. Sparse rewards can self-train dia-
logue agents, arXiv preprint arXiv:2409.04617, 2024. URL https://arxiv.org/abs/2409.
04617v2.
[580] Pak Kin Lau and Stuart Michael McManus. Mining asymmetric intertextuality, arXiv preprint
arXiv:2410.15145, 2024. URL https://arxiv.org/abs/2410.15145v1.
[581] Hung Le, T. Tran, and S. Venkatesh. Self-attentive associative memory. International Conference on
Machine Learning, 2020.
[582] Dohyun Lee, Seungil Chad Lee, Chanwoo Yang, Yujin Baek, and Jaegul Choo. Exploring in-context
example generation for machine translation, arXiv preprint arXiv:2506.00507, 2025. URL https:
//arxiv.org/abs/2506.00507v1.
[583] Dongyub Lee, Eunhwan Park, Hodong Lee, and Heuiseok Lim. Ask, assess, and refine: Rectifying
factual consistency and hallucination in llms with metric-guided feedback learning. Conference of the
European Chapter of the Association for Computational Linguistics, 2024.
[584] Eunhae Lee. Towards ethical personal ai applications: Practical considerations for ai assistants with
long-term memory, arXiv preprint arXiv:2409.11192, 2024. URL https://arxiv.org/abs/
2409.11192v1.
[585] Gibbeum Lee, Volker Hartmann, Jongho Park, Dimitris Papailiopoulos, and Kangwook Lee. Prompted
llms as chatbot modules for long open-domain conversation. In Findings of the Association for
Computational Linguistics: ACL 2023. Association for Computational Linguistics, 2023. doi: 10.18653/
v1/2023.findings-acl.277. URL http://dx.doi.org/10.18653/v1/2023.findings-acl.
277.
[586] Heejun Lee, Geon Park, Youngwan Lee, Jina Kim, Wonyoung Jeong, Myeongjae Jeon, and Sung Ju
Hwang. A training-free sub-quadratic cost transformer model serving framework with hierarchically
pruned attention, arXiv preprint arXiv:2406.09827, 2024. URL https://arxiv.org/abs/2406.
09827v3.
[587] Ho-Jun Lee, Junho Kim, Hyunjun Kim, and Yonghyun Ro. Refocus: Reinforcement-guided frame
optimization for contextual understanding, arXiv preprint arXiv:2506.01274v1, 2025. URL https:
//arxiv.org/abs/2506.01274v1.
[588] Jonathan Lee, Annie Xie, Aldo Pacchiano, Yash Chandak, Chelsea Finn, Ofir Nachum, and E. Brunskill.
Supervised pretraining can learn in-context reinforcement learning. Neural Information Processing
Systems, 2023.
[589] Namkyeong Lee, E. Brouwer, Ehsan Hajiramezanali, Chanyoung Park, and Gabriele Scalia. Rag-
enhanced collaborative llm agents for drug discovery, arXiv preprint arXiv:2502.17506, 2025. URL
https://arxiv.org/abs/2502.17506v2.
[590] Shinbok Lee, Gaeun Seo, Daniel Lee, Byeongil Ko, Sunghee Jung, and M. Shin. Functionchat-bench:
Comprehensive evaluation of language models’ generative capabilities in korean tool-use dialogs.
arXiv preprint, 2024.
[591] Younghun Lee, Sungchul Kim, Ryan A. Rossi, Tong Yu, and Xiang Chen. Learning to reduce:
Towards improving performance of large language models on structured data, arXiv preprint
arXiv:2407.02750, 2024. URL https://arxiv.org/abs/2407.02750v1.
[592] Younghun Lee, Sungchul Kim, Tong Yu, Ryan A. Rossi, and Xiang Chen. Learning to reduce:
Optimal representations of structured data in prompting large language models, arXiv preprint
arXiv:2402.14195, 2024. URL https://arxiv.org/abs/2402.14195v1.
[593] Yu-Ting Lee, Hui-Ying Shih, Fu-Chieh Chang, and Pei-Yuan Wu. An explanation of intrinsic self-
correction via linear representations and latent concepts, arXiv preprint arXiv:2505.11924, 2025.
URL https://arxiv.org/abs/2505.11924v1.
[594] Melissa Lehman and Kenneth J. Malmberg. A buffer model of memory encoding and temporal
correlations in retrieval. Psychology Review, 2013.
[595] Yiming Lei, Zhizheng Yang, Zeming Liu, Haitao Leng, Shaoguo Liu, Tingting Gao, Qingjie Liu,
and Yunhong Wang. Contextqformer: A new context modeling method for multi-turn multi-modal
conversations, arXiv preprint arXiv:2505.23121v1, 2025. URL https://arxiv.org/abs/2505.
23121v1.
[596] Brian Lester, Rami Al-Rfou, and Noah Constant. The power of scale for parameter-efficient prompt
tuning. Conference on Empirical Methods in Natural Language Processing, 2021.
[597] Patrick Lewis, Ethan Perez, Aleksandara Piktus, F. Petroni, Vladimir Karpukhin, Naman Goyal,
Heinrich Kuttler, M. Lewis, Wen tau Yih, Tim Rocktäschel, Sebastian Riedel, and Douwe Kiela.
Retrieval-augmented generation for knowledge-intensive nlp tasks. Neural Information Processing
Systems, 2020.
[598] Bohan Li, Yutai Hou, and Wanxiang Che. Data augmentation approaches in natural language
processing: A survey. AI Open, 2021.
[599] Chaozhuo Li, Bochen Pang, Yuming Liu, Hao Sun, Zheng Liu, Xing Xie, Tianqi Yang, Yanling Cui,
Liangjie Zhang, and Qi Zhang. Adsgnn: Behavior-graph augmented relevance modeling in sponsored
search. Annual International ACM SIGIR Conference on Research and Development in Information
Retrieval, 2021.
[600] Chengpeng Li, Zhengyang Tang, Ziniu Li, Mingfeng Xue, Keqin Bao, Tian Ding, Ruoyu Sun, Benyou
Wang, Xiang Wang, Junyang Lin, and Dayiheng Liu. Cort: Code-integrated reasoning within thinking,
arXiv preprint arXiv:2506.09820, 2025. URL https://arxiv.org/abs/2506.09820v2.
[601] Chengshu Li, Jacky Liang, Andy Zeng, Xinyun Chen, Karol Hausman, Dorsa Sadigh, Sergey Levine,
Fei-Fei Li, Fei Xia, and Brian Ichter. Chain of code: Reasoning with a language model-augmented
code emulator. International Conference on Machine Learning, 2023.
[602] Chuanhao Li, Runhan Yang, Tiankai Li, Milad Bafarassat, Kourosh Sharifi, Dirk Bergemann, and Zhuo-
ran Yang. Stride: A tool-assisted llm agent framework for strategic and interactive decision-making,
arXiv preprint arXiv:2405.16376, 2024. URL https://arxiv.org/abs/2405.16376v2.
[603] Daliang Li, Ankit Singh Rawat, Manzil Zaheer, Xin Wang, Michal Lukasik, Andreas Veit, Felix
Yu, and Sanjiv Kumar. Large language models with controllable working memory, arXiv preprint
arXiv:2211.05110, 2022. URL https://arxiv.org/abs/2211.05110.
[604] Daniel Li and Lincoln Murr. Humaneval on latest gpt models - 2024. arXiv preprint, 2024.
[605] Fu Li, Xueying Wang, Bin Li, Yunlong Wu, Yanzhen Wang, and Xiaodong Yi. A study on training and
developing large language models for behavior tree generation, arXiv preprint arXiv:2401.08089,
2024. URL https://arxiv.org/abs/2401.08089v1.
[606] Guohao Li, Hasan Abed Al Kader Hammoud, Hani Itani, Dmitrii Khizbullin, and Bernard Ghanem.
Camel: Communicative agents for “mind” exploration of large language model society. In Thirty-
seventh Conference on Neural Information Processing Systems, 2023.
[607] Guozheng Li, Peng Wang, Jiajun Liu, Yikai Guo, Ke Ji, Ziyu Shang, and Zijie Xu. Meta in-context
learning makes large language models better zero and few-shot relation extractors. International
Joint Conference on Artificial Intelligence, 2024.
[608] Haoyang Li, Jing Zhang, Cuiping Li, and Hong Chen. Resdsql: Decoupling schema linking and
skeleton parsing for text-to-sql. AAAI Conference on Artificial Intelligence, 2023.
[609] Jia Li, Ge Li, Yongming Li, and Zhi Jin. Structured chain-of-thought prompting for code generation.
ACM Transactions on Software Engineering and Methodology, 2023.
[610] Jia Li, Xiangguo Sun, Yuhan Li, Zhixun Li, Hong Cheng, and Jeffrey Xu Yu. Graph intelligence with
large language models and prompt learning. Knowledge Discovery and Data Mining, 2024.
[611] Jiahao Nick Li, Yan Xu, Tovi Grossman, Stephanie Santosa, and Michelle Li. Omniactions: Predicting
digital actions in response to real-world multimodal sensory inputs with llms. International Conference
on Human Factors in Computing Systems, 2024.
[612] Jiarui Li, Ye Yuan, and Zehua Zhang. Enhancing llm factual accuracy with rag to counter hal-
lucinations: A case study on domain-specific queries in private knowledge-bases, arXiv preprint
arXiv:2403.10446, 2024. URL https://arxiv.org/abs/2403.10446v1.
[613] Juncheng Li, Kaihang Pan, Zhiqi Ge, Minghe Gao, Hanwang Zhang, Wei Ji, Wenqiao Zhang, Tat
seng Chua, Siliang Tang, and Yueting Zhuang. Fine-tuning multimodal llms to follow zero-shot
demonstrative instructions. International Conference on Learning Representations, 2023.
[614] Junnan Li, Ramprasaath R. Selvaraju, Akhilesh Deepak Gotmare, Shafiq R. Joty, Caiming Xiong, and
S. Hoi. Align before fuse: Vision and language representation learning with momentum distillation.
Neural Information Processing Systems, 2021.
[615] Junnan Li, Dongxu Li, S. Savarese, and Steven C. H. Hoi. Blip-2: Bootstrapping language-image
pre-training with frozen image encoders and large language models. International Conference on
Machine Learning, 2023.
[616] Kun Li, Tianhua Zhang, Yunxiang Li, Hongyin Luo, Abdalla Moustafa, Xixin Wu, James Glass, and
Helen M. Meng. Generate, discriminate, evolve: Enhancing context faithfulness via fine-grained
sentence-level self-evolution, arXiv preprint arXiv:2503.01695, 2025. URL https://arxiv.org/
abs/2503.01695v1.
[617] Kunchang Li, Yinan He, Yi Wang, Yizhuo Li, Wen Wang, Ping Luo, Yali Wang, Limin Wang, and
Yu Qiao. Videochat: Chat-centric video understanding, arXiv preprint arXiv:2305.06355v2, 2023.
URL https://arxiv.org/abs/2305.06355v2.
[618] M Li, Y Zhao, B Yu, F Song, H Li, H Yu, and Z Li. . . . Api-bank: A comprehensive benchmark for
tool-augmented llms. 2023. URL https://arxiv.org/abs/2304.08244.
[619] Michelle M. Li, Ben Y. Reis, Adam Rodman, Tianxi Cai, Noa Dagan, Ran D. Balicer, Joseph Loscalzo,
Isaac S. Kohane, and M. Zitnik. One patient, many contexts: Scaling medical ai through contextual
intelligence, arXiv preprint arXiv:2506.10157, 2025. URL https://arxiv.org/abs/2506.
10157v1.
[620] Ming Li, Keyu Chen, Ziqian Bi, Ming Liu, Benji Peng, Qian Niu, Junyu Liu, Jinlang Wang, Sen
Zhang, Xuanhe Pan, Jiawei Xu, and Pohsun Feng. Surveying the mllm landscape: A meta-review of
current surveys, arXiv preprint arXiv:2409.18991, 2024. URL https://arxiv.org/abs/2409.
18991v1.
[621] Minghao Li, Feifan Song, Yu Bowen, Haiyang Yu, Zhoujun Li, Fei Huang, and Yongbin Li. Api-bank:
A comprehensive benchmark for tool-augmented llms. Conference on Empirical Methods in Natural
Language Processing, 2023.
[622] Qiaomu Li and Ying Xie. From glue-code to protocols: A critical analysis of a2a and mcp integration
for scalable agent systems, arXiv preprint arXiv:2505.03864, 2025. URL https://arxiv.org/
abs/2505.03864v1.
[623] Rongsheng Li, Jin Xu, Zhixiong Cao, Hai-Tao Zheng, and Hong-Gee Kim. Extending context window
in large language models with segmented base adjustment for rotary position embeddings. Applied
Sciences, 2024.
[624] Shuaike Li, Kai Zhang, Qi Liu, and Enhong Chen. Mindbridge: Scalable and cross-model knowledge
editing via memory-augmented modality, arXiv preprint arXiv:2503.02701v1, 2025. URL https:
//arxiv.org/abs/2503.02701v1.
[625] Shuaiyi Li, Zhisong Zhang, Yang Deng, Chenlong Deng, Tianqing Fang, Hongming Zhang, Haitao
Mi, Dong Yu, and Wai Lam. Incomes: Integrating compression and selection mechanisms into llms
for efficient model editing, arXiv preprint arXiv:2505.22156, 2025. URL https://arxiv.org/
abs/2505.22156v1.
[626] Siheng Li, Cheng Yang, Zesen Cheng, Lemao Liu, Mo Yu, Yujiu Yang, and Wai Lam. Large language
models can self-improve in long-context reasoning, arXiv preprint arXiv:2411.08147, 2024. URL
https://arxiv.org/abs/2411.08147v1.
[627] X Li, H Zou, and P Liu. Torl: Scaling tool-integrated rl. 2025. URL https://arxiv.org/abs/
2503.23383.
[628] Xiaopeng Li, Pengyue Jia, Derong Xu, Yi Wen, Yingyi Zhang, Wenlin Zhang, Wanyu Wang, Yichao
Wang, Zhaochen Du, Xiangyang Li, Yong Liu, Huifeng Guo, Ruiming Tang, and Xiangyu Zhao.
A survey of personalization: From rag to agent, arXiv preprint arXiv:2504.10147, 2025. URL
https://arxiv.org/abs/2504.10147v1.
[629] Xiaoxi Li, Jiajie Jin, Guanting Dong, Hongjin Qian, Yutao Zhu, Yongkang Wu, Ji-Rong Wen, and
Zhicheng Dou. Webthinker: Empowering large reasoning models with deep research capability,
arXiv preprint arXiv:2504.21776, 2025. URL https://arxiv.org/abs/2504.21776v1.
[630] Xin Li, Qizhi Chu, Yubin Chen, Yang Liu, Yaoqi Liu, Zekai Yu, Weize Chen, Cheng Qian, Chuan Shi,
and Cheng Yang. Graphteam: Facilitating large language model-based graph analysis via multi-agent
collaboration, arXiv preprint arXiv:2410.18032v4, 2024. URL https://arxiv.org/abs/2410.
18032v4.
[631] Xinyi Li, Sai Wang, Siqi Zeng, Yu Wu, and Yi Yang. A survey on llm-based multi-agent systems:
workflow, infrastructure, and challenges. Vicinagearth, 2024.
[632] Xinze Li, Zhenghao Liu, Chenyan Xiong, Shi Yu, Yu Gu, Zhiyuan Liu, and Ge Yu. Structure-aware
language model pretraining improves dense retrieval on structured data. Annual Meeting of the
Association for Computational Linguistics, 2023.
[633] Yang Li, Jiacong He, Xiaoxia Zhou, Yuan Zhang, and Jason Baldridge. Mapping natural language
instructions to mobile ui action sequences. Annual Meeting of the Association for Computational
Linguistics, 2020.
[634] Yinghao Li, R. Ramprasad, and Chao Zhang. A simple but effective approach to improve structured
language model output for information extraction. Conference on Empirical Methods in Natural
Language Processing, 2024.
[635] Yuan Li, Yixuan Zhang, and Lichao Sun. Metaagents: Simulating interactions of human behav-
iors for llm-based task-oriented coordination via collaborative generative agents, arXiv preprint
arXiv:2310.06500, 2023. URL https://arxiv.org/abs/2310.06500.
[636] Yucheng Li. Unlocking context constraints of llms: Enhancing context efficiency of llms with
self-information-based content filtering, arXiv preprint arXiv:2304.12102, 2023. URL https:
//arxiv.org/abs/2304.12102v1.
[637] Yucheng Li, Bo Dong, Chenghua Lin, and Frank Guerin. Compressing context to enhance inference
efficiency of large language models. Conference on Empirical Methods in Natural Language Processing,
2023.
[638] Zhaoxin Li, Xiaoming Zhang, Haifeng Zhang, and Chengxiang Liu. Refining interactions: Enhancing
anisotropy in graph neural networks with language semantics, arXiv preprint arXiv:2504.01429,
2025. URL https://arxiv.org/abs/2504.01429v1.
[639] Zhecheng Li, Yiwei Wang, Bryan Hooi, Yujun Cai, Naifan Cheung, Nanyun Peng, and Kai-Wei Chang.
Think carefully and check again! meta-generation unlocking llms for low-resource cross-lingual
summarization. 2024.
[640] Zhecheng Li, Yiwei Wang, Bryan Hooi, Yujun Cai, Nanyun Peng, and Kai-Wei Chang. Drs: Deep
question reformulation with structured output. In Association for Computational Linguistics ACL,
2025., 2024.
[641] Zhecheng Li, Yiwei Wang, Bryan Hooi, Yujun Cai, Zhen Xiong, Nanyun Peng, and Kai-Wei Chang.
Vulnerability of llms to vertically aligned text manipulations. In Association for Computational
Linguistics ACL, 2025., 2024.
[642] Zhecheng Li, Guoxian Song, Yujun Cai, Zhen Xiong, Junsong Yuan, and Yiwei Wang. Texture or
semantics? vision-language models get lost in font recognition. In Conference on Language Modeling
COLM, 2025., 2025.
[643] Zhiyu Li, Shichao Song, Hanyu Wang, Simin Niu, Ding Chen, Jiawei Yang, Chenyang Xi, Huayi Lai,
Jihao Zhao, Yezhaohui Wang, Junpeng Ren, Zehao Lin, Jiahao Huo, Tianyi Chen, Kai Chen, Ke-Rong
Li, Zhiqiang Yin, Qingchen Yu, Bo Tang, Hongkang Yang, Zhiyang Xu, and Feiyu Xiong. Memos: An
operating system for memory-augmented generation (mag) in large language models, arXiv preprint
arXiv:2505.22101, 2025. URL https://arxiv.org/abs/2505.22101v1.
[644] Zhong-Zhi Li, Ming-Liang Zhang, Fei Yin, and Cheng-Lin Liu. Lans: A layout-aware neural solver for
plane geometry problem. 2023.
[645] Zhong-Zhi Li, Ming-Liang Zhang, Fei Yin, Zhi-Long Ji, Jin-Feng Bai, Zhen-Ru Pan, Fan-Hu Zeng,
Jian Xu, Jia-Xin Zhang, and Cheng-Lin Liu. Cmmath: A chinese multi-modal math skill evaluation
benchmark for foundation models. 2024.
[646] Zhong-Zhi Li, Duzhen Zhang, Ming-Liang Zhang, Jiaxin Zhang, Zengyan Liu, Yuxuan Yao, Haotian
Xu, Junhao Zheng, Pei-Jie Wang, Xiuyi Chen, et al. From system 1 to system 2: A survey of reasoning
large language models. 2025.
[647] Zhuoqun Li, Xuanang Chen, Haiyang Yu, Hongyu Lin, Yaojie Lu, Qiaoyu Tang, Fei Huang, Xianpei
Han, Le Sun, and Yongbin Li. Structrag: Boosting knowledge intensive reasoning of llms via inference-
time hybrid information structurization. International Conference on Learning Representations, 2024.
[648] Zinuo Li, Xian Zhang, Yongxin Guo, Mohammed Bennamoun, F. Boussaid, Girish Dwivedi, Luqi Gong,
and Qiuhong Ke. Watch and listen: Understanding audio-visual-speech moments with multimodal
llm, arXiv preprint arXiv:2505.18110v2, 2025. URL https://arxiv.org/abs/2505.18110v2.
[649] Zixuan Li, Jing Xiong, Fanghua Ye, Chuanyang Zheng, Xun Wu, Jianqiao Lu, Zhongwei Wan, Xiaodan
Liang, Chengming Li, Zhenan Sun, Lingpeng Kong, and Ngai Wong. Uncertaintyrag: Span-level
uncertainty enhanced long-context modeling for retrieval-augmented generation, arXiv preprint
arXiv:2410.02719, 2024. URL https://arxiv.org/abs/2410.02719v1.
[650] Zonglin Li, Ruiqi Guo, and Surinder Kumar. Decoupled context processing for context augmented
language modeling. Neural Information Processing Systems, 2022.
[651] Zongqian Li, Yinhong Liu, Yixuan Su, and Nigel Collier. Prompt compression for large language
models: A survey. North American Chapter of the Association for Computational Linguistics, 2024.
[652] Wen li Yu and Junfeng Zhao. Quantum multi-agent reinforcement learning as an emerging ai
technology: A survey and future directions. International Conferences on Computing Advancements,
2023.
[653] Guannan Liang and Qianqian Tong. Llm-powered ai agent systems and their applications in industry,
arXiv preprint arXiv:2505.16120, 2025. URL https://arxiv.org/abs/2505.16120v1.
[654] Jintao Liang, Gang Su, Huifeng Lin, You Wu, Rui Zhao, and Ziyue Li. Reasoning rag via system 1 or
system 2: A survey on reasoning agentic retrieval-augmented generation for industry challenges,
arXiv preprint arXiv:2506.10408, 2025. URL https://arxiv.org/abs/2506.10408v1.
[655] Xinnian Liang, Bing Wang, Huijia Huang, Shuangzhi Wu, Peihao Wu, Lu Lu, Zejun Ma, and Zhoujun
Li. Scm: Enhancing large language model with self-controlled memory framework, arXiv preprint
arXiv:2304.13343, 2023. URL https://arxiv.org/abs/2304.13343v4.
[656] Xuechen Liang, Meiling Tao, Yinghui Xia, Tianyu Shi, Jun Wang, and Jingsong Yang. Self-evolving
agents with reflective and memory-augmented abilities, arXiv preprint arXiv:2409.00872, 2024.
URL https://arxiv.org/abs/2409.00872v2.
[657] Xuechen Liang, Meiling Tao, Yinghui Xia, Jianhui Wang, Kun Li, Yijin Wang, Jingsong Yang, Tianyu
Shi, Yuantao Wang, Miao Zhang, and Xueqian Wang. Mars: Memory-enhanced agents with reflective
self-improvement, arXiv preprint arXiv:2503.19271, 2025. URL https://arxiv.org/abs/2503.
19271v2.
[658] Yanbiao Liang, Huihong Shi, Haikuo Shao, and Zhongfeng Wang. Accllm: Accelerating long-context
llm inference via algorithm-hardware co-design, arXiv preprint arXiv:2505.03745, 2025. URL
https://arxiv.org/abs/2505.03745v1.
[659] Yaobo Liang, Chenfei Wu, Ting Song, Wenshan Wu, Yan Xia, Yu Liu, Yangyiwen Ou, Shuai Lu, Lei Ji,
Shaoguang Mao, Yun Wang, Linjun Shou, Ming Gong, and Nan Duan. Taskmatrix.ai: Completing
tasks by connecting foundation models with millions of apis. Intelligent Computing, 2023.
[660] Yiming Liang, Ge Zhang, Xingwei Qu, Tianyu Zheng, Jiawei Guo, Xinrun Du, Zhenzhu Yang, Jiaheng
Liu, Chenghua Lin, Lei Ma, Wenhao Huang, and Jiajun Zhang. I-sheep: Self-alignment of llm from
scratch through an iterative self-enhancement paradigm. arXiv preprint, 2024.
[661] Bingli Liao and Danilo Vasconcellos Vargas. Beyond kv caching: Shared attention for efficient llms.
Neurocomputing, 2024.
[662] Xiaoxuan Liao, Binrong Zhu, Jacky He, Guiran Liu, Hongye Zheng, and Jia Gao. A fine-tuning
approach for t5 using knowledge graphs to address complex tasks, arXiv preprint arXiv:2502.16484,
2025. URL https://arxiv.org/abs/2502.16484v1.
[663] David Lillis. Internalising interaction protocols as first-class programming elements in multi agent sys-
tems, arXiv preprint arXiv:1711.02634, 2017. URL https://arxiv.org/abs/1711.02634v1.
[664] Bill Yuchen Lin, Xinyue Chen, Jamin Chen, and Xiang Ren. Kagnet: Knowledge-aware graph networks
for commonsense reasoning. Conference on Empirical Methods in Natural Language Processing, 2019.
[665] Bill Yuchen Lin, Seyeon Lee, Xiaoyang Qiao, and Xiang Ren. Common sense beyond english:
Evaluating and improving multilingual language models for commonsense reasoning. Annual
Meeting of the Association for Computational Linguistics, 2021.
[666] Bin Lin, Chen Zhang, Tao Peng, Hanyu Zhao, Wencong Xiao, Minmin Sun, Anmin Liu, Zhipeng
Zhang, Lanbo Li, Xiafei Qiu, Shen Li, Zhigang Ji, Tao Xie, Yong Li, and Wei Lin. Infinite-llm:
Efficient llm service for long context with distattention and distributed kvcache, arXiv preprint
arXiv:2401.02669, 2024. URL https://arxiv.org/abs/2401.02669.
[667] Jianhao Lin, Lexuan Sun, and Yixin Yan. Simulating macroeconomic expectations using llm agents,
arXiv preprint arXiv:2505.17648, 2025. URL https://arxiv.org/abs/2505.17648v2.
[668] Lei Lin, Jiayi Fu, Pengli Liu, Qingyang Li, Yan Gong, Junchen Wan, Fuzheng Zhang, Zhongyuan
Wang, Di Zhang, and Kun Gai. Just ask one more time! self-agreement improves reasoning of
language models in (almost) all scenarios. Annual Meeting of the Association for Computational
Linguistics, 2023.
[669] Matthieu Lin, Jenny Sheng, Andrew Zhao, Shenzhi Wang, Yang Yue, Yiran Wu, Huan Liu, Jun Liu, Gao
Huang, and Yong-Jin Liu. Training of scaffolded language models with language supervision: A survey,
arXiv preprint arXiv:2410.16392, 2024. URL https://arxiv.org/abs/2410.16392v2.
[670] Qiqiang Lin, Muning Wen, Qiuying Peng, Guanyu Nie, Junwei Liao, Jun Wang, Xiaoyun Mo, Jiamu
Zhou, Cheng Cheng, Yin Zhao, and Weinan Zhang. Hammer: Robust function-calling for on-
device language models via function masking, arXiv preprint arXiv:2410.04587, 2024. URL https:
//arxiv.org/abs/2410.04587v2.
[671] Yu-Chen Lin, Akhilesh Kumar, Norman Chang, Wen-Liang Zhang, Muhammad Zakir, Rucha Apte,
Haiyang He, Chao Wang, and Jyh-Shing Roger Jang. Novel preprocessing technique for data
embedding in engineering code generation using large language model. 2024 IEEE LLM Aided Design
Workshop (LAD), 2023.
[672] Yu-Hsuan Lin, Qian-Hui Chen, Yi-Jie Cheng, Jia-Ren Zhang, Yi-Hung Liu, Liang-Yu Hsia, and
Yun-Nung Chen. Llm inference enhanced by external knowledge: A survey, arXiv preprint
arXiv:2505.24377, 2025. URL https://arxiv.org/abs/2505.24377v1.
[673] Jack W Lindsey and Ashok Litwin-Kumar. Selective consolidation of learning and memory via
recall-gated plasticity. bioRxiv, 2024.
[674] Tal Linzen, Emmanuel Dupoux, and Yoav Goldberg. Assessing the ability of lstms to learn syntax-
sensitive dependencies. Transactions of the Association for Computational Linguistics, 2016.
[675] Gili Lior, Yuval Shalev, Gabriel Stanovsky, and Ariel Goldstein. Computation or weight adaptation?
rethinking the role of plasticity in learning. bioRxiv, 2024.
[676] Bingyang Liu, Haoyi Zhang, Xiaohan Gao, Zichen Kong, Xiyuan Tang, Yibo Lin, Runsheng Wang,
and Ru Huang. Layoutcopilot: An llm-powered multi-agent collaborative framework for interactive
analog layout design. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems,
2024.
[677] E. Liu, Kelvin Guu, Panupong Pasupat, Tianlin Shi, and Percy Liang. Reinforcement learning on web
interfaces using workflow-guided exploration. International Conference on Learning Representations,
2018.
[678] Guang-Da Liu, Haitao Mao, Bochuan Cao, Zhiyu Xue, K. Johnson, Jiliang Tang, and Rongrong Wang.
On the intrinsic self-correction capability of llms: Uncertainty and latent concept, arXiv preprint
arXiv:2406.02378, 2024. URL https://arxiv.org/abs/2406.02378v2.
[679] Guangyi Liu, Yongqi Zhang, Yong Li, and Quanming Yao. Dual reasoning: A gnn-llm collaborative
framework for knowledge graph question answering, arXiv preprint arXiv:2406.01145, 2024. URL
https://arxiv.org/abs/2406.01145v2.
[680] Guangyi Liu, Pengxiang Zhao, Liang Liu, Yaxuan Guo, Han Xiao, Weifeng Lin, Yuxiang Chai, Yue Han,
Shuai Ren, Hao Wang, Xiaoyu Liang, Wenhao Wang, Tianze Wu, Linghao Li, Guanjing Xiong, Yong Liu,
and Hongsheng Li. Llm-powered gui agents in phone automation: Surveying progress and prospects,
arXiv preprint arXiv:2504.19838, 2025. URL https://arxiv.org/abs/2504.19838v2.
[681] Hanchao Liu, Rong-Zhi Li, Weimin Xiong, Ziyu Zhou, and Wei Peng. Workteam: Constructing
workflows from natural language with multi-agents. North American Chapter of the Association for
Computational Linguistics, 2025.
[682] Hao Liu, Matei Zaharia, and Pieter Abbeel. Ring attention with blockwise transformers for near-
infinite context. International Conference on Learning Representations, 2023.
[683] Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. Visual instruction tuning. Neural
Information Processing Systems, 2023.
[684] Jie Liu, Pan Zhou, Yingjun Du, Ah-Hwee Tan, Cees G. M. Snoek, J. Sonke, and E. Gavves. Capo: Co-
operative plan optimization for efficient embodied multi-agent cooperation. International Conference
on Learning Representations, 2024.
[685] Jun Liu, Ke Yu, Keliang Chen, Ke Li, Yuxinyue Qian, Xiaolian Guo, Haozhe Song, and Yinming Li.
Acps: Agent collaboration protocols for the internet of agents, arXiv preprint arXiv:2505.13523,
2025. URL https://arxiv.org/abs/2505.13523v1.
[686] Junwei Liu, Kaixin Wang, Yixuan Chen, Xin Peng, Zhenpeng Chen, Lingming Zhang, and Yil-
ing Lou. Large language model-based agents for software engineering: A survey, arXiv preprint
arXiv:2409.02977, 2024. URL https://arxiv.org/abs/2409.02977v1.
[687] Kai Liu, Zhihang Fu, Chao Chen, Wei Zhang, Rongxin Jiang, Fan Zhou, Yao-Shen Chen, Yue Wu,
and Jieping Ye. Enhancing llm’s cognition via structurization. Neural Information Processing Systems,
2024.
[688] Lei Liu, Xiaoyan Yang, Yue Shen, Binbin Hu, Zhiqiang Zhang, Jinjie Gu, and Guannan Zhang.
Think-in-memory: Recalling and post-thinking enable llms with long-term memory. arXiv preprint,
2023.
[689] Lei Liu, Xiaoyan Yang, Yue Shen, Binbin Hu, Zhiqiang Zhang, Jinjie Gu, and Guannan Zhang.
Think-in-memory: Recalling and post-thinking enable llms with long-term memory, arXiv preprint
arXiv:2311.08719, 2023. URL https://arxiv.org/abs/2311.08719.
[690] Na Liu, Liangyu Chen, Xiaoyu Tian, Wei Zou, Kaijiang Chen, and Ming Cui. From llm to conversational
agent: A memory enhanced architecture with fine-tuning of large language models, arXiv preprint
arXiv:2401.02777, 2024. URL https://arxiv.org/abs/2401.02777v2.
[691] Nelson F. Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, F. Petroni, and Percy
Liang. Lost in the middle: How language models use long contexts. Transactions of the Association
for Computational Linguistics, 2023.
[692] Pei Liu, Xin Liu, Ruoyu Yao, Junming Liu, Siyuan Meng, Ding Wang, and Jun Ma. Hm-rag: Hierar-
chical multi-agent multimodal retrieval augmented generation. arXiv preprint, 2025.
[693] Shicheng Liu, Jialiang Xu, Wesley Tjangnaka, Sina J. Semnani, Chen Jie Yu, Gui D’avid, and Monica S.
Lam. Suql: Conversational search over structured and unstructured data with large language models.
NAACL-HLT, 2023.
[694] Shiyu Liu, Yucheng Han, Peng Xing, Fukun Yin, Rui Wang, Wei Cheng, Jiaqi Liao, Yingming Wang,
Honghao Fu, Chunrui Han, et al. Step1x-edit: A practical framework for general image editing.
2025.
[695] W Liu, X Huang, X Zeng, X Hao, S Yu, and D Li. . . . Toolace: Winning the points of llm function
calling. 2024. URL https://arxiv.org/abs/2409.00920.
[696] Weijie Liu, Peng Zhou, Zhe Zhao, Zhiruo Wang, Qi Ju, Haotang Deng, and Ping Wang. K-bert:
Enabling language representation with knowledge graph. AAAI Conference on Artificial Intelligence,
2019.
[697] Weiwen Liu, Xu Huang, Xingshan Zeng, Xinlong Hao, Shuai Yu, Dexun Li, Shuai Wang, Weinan Gan,
Zhengying Liu, Yuanqing Yu, Zezhong Wang, Yuxian Wang, Wu Ning, Yutai Hou, Bin Wang, Chuhan
Wu, Xinzhi Wang, Yong Liu, Yasheng Wang, Duyu Tang, Dandan Tu, Lifeng Shang, Xin Jiang, Ruiming
Tang, Defu Lian, Qun Liu, and Enhong Chen. Toolace: Winning the points of llm function calling,
arXiv preprint arXiv:2409.00920, 2024. URL https://arxiv.org/abs/2409.00920v1.
[698] Weiwen Liu, Jiarui Qin, Xu Huang, Xingshan Zeng, Yunjia Xi, Jianghao Lin, Chuhan Wu, Yasheng
Wang, Lifeng Shang, Ruiming Tang, Defu Lian, Yong Yu, and Weinan Zhang. The real barrier to llm
agent usability is agentic roi, arXiv preprint arXiv:2505.17767, 2025. URL https://arxiv.org/
abs/2505.17767v1.
[699] Wentao Liu, Hanglei Hu, Jie Zhou, Yuyang Ding, Junsong Li, Jiayi Zeng, Mengliang He, Qin Chen,
Bo Jiang, Aimin Zhou, and Liang He. Mathematical language models: A survey, arXiv preprint
arXiv:2312.07622, 2023. URL https://arxiv.org/abs/2312.07622v4.
[700] Wentao Liu, Ruohua Zhang, Aimin Zhou, Feng Gao, and JiaLi Liu. Echo: A large language model
with temporal episodic memory, arXiv preprint arXiv:2502.16090, 2025. URL https://arxiv.
org/abs/2502.16090v1.
[701] Xu Liu, S. Ramirez, Petti T. Pang, C. Puryear, A. Govindarajan, K. Deisseroth, and S. Tonegawa.
Optogenetic stimulation of a hippocampal engram activates fear memory recall. Nature, 2012.
[702] Yang Liu, Xiaobin Tian, Zequn Sun, and Wei Hu. Finetuning generative large language models
with discrimination instructions for knowledge graph completion. In International Semantic Web
Conference, 2024.
[703] Yue Liu, Jiaying Wu, Yufei He, Hongcheng Gao, Hongyu Chen, Baolong Bi, Jiaheng Zhang, Zhiqi
Huang, and Bryan Hooi. Efficient inference for large reasoning models: A survey, arXiv preprint
arXiv:2503.23077, 2025. URL https://arxiv.org/abs/2503.23077v2.
[704] Zijun Liu, Yanzhe Zhang, Peng Li, Yang Liu, and Diyi Yang. A dynamic llm-powered agent network for
task-oriented agent collaboration, arXiv preprint arXiv:2310.02170, 2023. URL https://arxiv.
org/abs/2310.02170v2.
[705] Zijun Liu, Zhennan Wan, Peng Li, Ming Yan, Ji Zhang, Fei Huang, and Yang Liu. Scaling external
knowledge input beyond context windows of llms via multi-agent collaboration, arXiv preprint
arXiv:2505.21471, 2025. URL https://arxiv.org/abs/2505.21471v1.
[706] Zinan Liu, Haoran Li, Jingyi Lu, Gaoyuan Ma, Xu Hong, Giovanni Iacca, Arvind Kumar, Shaojun
Tang, and Lin Wang. Nature’s insight: A novel framework and comprehensive analysis of agentic
reasoning through the lens of neuroscience. arXiv preprint, 2025.
[707] Zuxin Liu, Thai Hoang, Jianguo Zhang, Ming Zhu, Tian Lan, Shirley Kokane, Juntao Tan, Weiran
Yao, Zhiwei Liu, Yihao Feng, Rithesh Murthy, Liangwei Yang, Silvio Savarese, Juan Carlos Niebles,
Huan Wang, Shelby Heinecke, and Caiming Xiong. Apigen: Automated pipeline for generating
verifiable and diverse function-calling datasets. Neural Information Processing Systems, 2024.
[708] Leo S. Lo. The art and science of prompt engineering: A new literacy in the information age. Internet
Reference Services Quarterly, 2023.
[709] Joseph R. Loffredo and Suyeol Yun. Agent-enhanced large language models for researching political
institutions, arXiv preprint arXiv:2503.13524, 2025. URL https://arxiv.org/abs/2503.
13524v1.
[710] Jianqiao Lu, Wanjun Zhong, Wenyong Huang, Yufei Wang, Fei Mi, Baojun Wang, Weichao
Wang, Lifeng Shang, and Qun Liu. Self: Self-evolution with language feedback, arXiv preprint
arXiv:2310.00533, 2023. URL https://arxiv.org/abs/2310.00533v4.
[711] Junru Lu, Siyu An, Mingbao Lin, Gabriele Pergola, Yulan He, Di Yin, Xing Sun, and Yunsheng Wu.
Memochat: Tuning llms to use memos for consistent long-range open-domain conversation, arXiv
preprint arXiv:2308.08239, 2023. URL https://arxiv.org/abs/2308.08239.
[712] Junting Lu, Zhiyang Zhang, Fangkai Yang, Jue Zhang, Lu Wang, Chao Du, Qingwei Lin, Saravan
Rajmohan, Dongmei Zhang, and Qi Zhang. Axis: Efficient human-agent-computer interaction with
api-first llm-based agents, arXiv preprint arXiv:2409.17140, 2025. URL https://arxiv.org/
abs/2409.17140.
[713] Keer Lu, Xiaonan Nie, Zheng Liang, Da Pan, Shusen Zhang, Keshi Zhao, Weipeng Chen, Zenan
Zhou, Guosheng Dong, Bin Cui, and Wentao Zhang. Datasculpt: Crafting data landscapes for
long-context llms through multi-objective partitioning, arXiv preprint arXiv:2409.00997, 2024. URL
https://arxiv.org/abs/2409.00997v2.
[714] Liqiang Lu, Yicheng Jin, Hangrui Bi, Zizhang Luo, Peng Li, Tao Wang, and Yun Liang. Sanger: A
co-design framework for enabling sparse attention using reconfigurable architecture. Micro, 2021.
[715] Pan Lu, Baolin Peng, Hao Cheng, Michel Galley, Kai-Wei Chang, Y. Wu, Song-Chun Zhu, and Jianfeng
Gao. Chameleon: Plug-and-play compositional reasoning with large language models. Neural
Information Processing Systems, 2023.
[716] Y Lu, H Yu, and D Khashabi. Gear: Augmenting language models with generalizable and efficient
tool resolution. 2023. URL https://arxiv.org/abs/2307.08775.
[717] Yao Lu, Max Bartolo, Alastair Moore, Sebastian Riedel, and Pontus Stenetorp. Fantastically ordered
prompts and where to find them: Overcoming few-shot prompt order sensitivity. Annual Meeting of
the Association for Computational Linguistics, 2021.
[718] Yinquan Lu, H. Lu, Guirong Fu, and Qun Liu. Kelm: Knowledge enhanced pre-trained language repre-
sentations with message passing on hierarchical relational graphs, arXiv preprint arXiv:2109.04223,
2021. URL https://arxiv.org/abs/2109.04223v2.
[719] Elias Lumer, Anmol Gulati, V. K. Subbiah, Pradeep Honaganahalli Basavaraju, and James A. Burke.
Scalemcp: Dynamic and auto-synchronizing model context protocol tools for llm agents, arXiv
preprint arXiv:2505.06416, 2025. URL https://arxiv.org/abs/2505.06416v1.
[720] Cheng Luo, Jiawei Zhao, Zhuoming Chen, Beidi Chen, and Anima Anandkumar. Mini-sequence
transformer: Optimizing intermediate memory for long sequences training, arXiv preprint
arXiv:2407.15892, 2024. URL https://arxiv.org/abs/2407.15892v4.
[721] Feng Luo, Yu-Neng Chuang, Guanchu Wang, Hoang Anh Duy Le, Shaochen Zhong, Hongyi Liu,
Jiayi Yuan, Yang Sui, Vladimir Braverman, Vipin Chaudhary, and Xia Hu. Autol2s: Auto long-
short reasoning for efficient large language models, arXiv preprint arXiv:2505.22662, 2025. URL
https://arxiv.org/abs/2505.22662v1.
[722] Haitong Luo, Xuying Meng, Suhang Wang, Tianxiang Zhao, Fali Wang, Hanyun Cao, and Yujun
Zhang. Enhance graph alignment for large language models, arXiv preprint arXiv:2410.11370v1,
2024. URL https://arxiv.org/abs/2410.11370v1.
[723] Haoran Luo, E. Haihong, Guanting Chen, Yandan Zheng, Xiaobao Wu, Yikai Guo, Qika Lin, Yu Feng,
Zemin Kuang, Meina Song, Yifan Zhu, and Anh Tuan Luu. Hypergraphrag: Retrieval-augmented
generation with hypergraph-structured knowledge representation. arXiv preprint, 2025.
[724] Haotian Luo, Li Shen, Haiying He, Yibo Wang, Shiwei Liu, Wei Li, Naiqiang Tan, Xiaochun Cao, and
Dacheng Tao. O1-pruner: Length-harmonizing fine-tuning for o1-like reasoning pruning. arXiv
preprint, 2025.
[725] Junyu Luo, Weizhi Zhang, Ye Yuan, Yusheng Zhao, Junwei Yang, Yiyang Gu, Bohan Wu, Binqi
Chen, Ziyue Qiao, Qingqing Long, Rongcheng Tu, Xiaoming Luo, Wei Ju, Zhiping Xiao, Yifan Wang,
Mengxue Xiao, Chenwu Liu, Jingyang Yuan, Shichang Zhang, Yiqiao Jin, Fan Zhang, Xianhong
Wu, Hanqing Zhao, Dacheng Tao, Philip S. Yu, and Ming Zhang. Large language model agent: A
survey on methodology, applications and challenges, arXiv preprint arXiv:2503.21460, 2025. URL
https://arxiv.org/abs/2503.21460v1.
[726] Linhao Luo, Yuan-Fang Li, Gholamreza Haffari, and Shirui Pan. Reasoning on graphs: Faithful and
interpretable large language model reasoning. International Conference on Learning Representations,
2023.
[727] Renjie Luo, Jiaxi Li, Chen Huang, and Wei Lu. Through the valley: Path to effective long cot training
for small language models, arXiv preprint arXiv:2506.07712, 2025. URL https://arxiv.org/
abs/2506.07712v1.
[728] Xindi Luo, Zequn Sun, Jing Zhao, Zhe Zhao, and Wei Hu. Knowla: Enhancing parameter-efficient
finetuning with knowledgeable adaptation. In Proceedings of the 2024 Conference of the North
American Chapter of the Association for Computational Linguistics, 2024.
[729] Yifan Luo, Yiming Tang, Chengfeng Shen, Zhennan Zhou, and Bin Dong. Prompt engineering
through the lens of optimal control. Journal of Machine Learning, 2023.
[730] Panagiotis Lymperopoulos and Vasanth Sarathy. Tools in the loop: Quantifying uncertainty of llm
question answering systems that use tools, arXiv preprint arXiv:2505.16113, 2025. URL https:
//arxiv.org/abs/2505.16113v1.
[731] Yougang Lyu, Xiaoyu Zhang, Lingyong Yan, M. D. Rijke, Zhaochun Ren, and Xiuying Chen. Deepshop:
A benchmark for deep research shopping agents, arXiv preprint arXiv:2506.02839, 2025. URL
https://arxiv.org/abs/2506.02839v1.
[732] Jianxiang Ma. Research on the role of llm in multi-agent systems: A survey. Applied and Computational
Engineering, 2024.
[733] Jie Ma, Zhitao Gao, Qianyi Chai, Wangchun Sun, Pinghui Wang, Hongbin Pei, Jing Tao, Lingyun
Song, Jun Liu, Chen Zhang, and Li zhen Cui. Debate on graph: a flexible and reliable reasoning
framework for large language models, arXiv preprint arXiv:2409.03155, 2024. URL https://
arxiv.org/abs/2409.03155v1.
[734] Qun Ma, Xiao Xue, Deyu Zhou, Xiangning Yu, Donghua Liu, Xuwen Zhang, Zihan Zhao, Yifan
Shen, Peilin Ji, Juanjuan Li, Gang Wang, and Wanpeng Ma. Computational experiments meet large
language model based agents: A survey and perspective, arXiv preprint arXiv:2402.00262, 2024.
URL https://arxiv.org/abs/2402.00262v1.
[735] Xin Ma, Yang Liu, Jingjing Liu, and Xiaoxu Ma. Mesa-extrapolation: A weave position encoding
method for enhanced extrapolation in llms. Neural Information Processing Systems, 2024.
[736] Xinbei Ma, Yeyun Gong, Pengcheng He, Hai Zhao, and Nan Duan. Query rewriting for retrieval-
augmented large language models, arXiv preprint arXiv:2305.14283, 2023. URL https://arxiv.
org/abs/2305.14283v3.
[737] Xuezhe Ma, Xiaomeng Yang, Wenhan Xiong, Beidi Chen, Lili Yu, Hao Zhang, Jonathan May, Luke S.
Zettlemoyer, Omer Levy, and Chunting Zhou. Megalodon: Efficient llm pretraining and inference
with unlimited context length. Neural Information Processing Systems, 2024.
[738] Y Ma, Z Gou, J Hao, R Xu, S Wang, and L Pan. . . . Sciagent: Tool-augmented language models for
scientific reasoning. 2024. URL https://arxiv.org/abs/2402.11451.
[739] Zhiyuan Ma, Zhenya Huang, Jiayu Liu, Minmao Wang, Hongke Zhao, and Xin Li. Automated creation
of reusable and diverse toolsets for enhancing llm reasoning. AAAI Conference on Artificial Intelligence,
2025.
[740] Aman Madaan, Niket Tandon, Peter Clark, and Yiming Yang. Memory-assisted prompt editing to
improve gpt-3 after deployment. Conference on Empirical Methods in Natural Language Processing,
2022.
[741] Aman Madaan, Niket Tandon, Prakhar Gupta, Skyler Hallinan, Luyu Gao, Sarah Wiegreffe, Uri Alon,
Nouha Dziri, Shrimai Prabhumoye, Yiming Yang, S. Welleck, Bodhisattwa Prasad Majumder, Shashank
Gupta, A. Yazdanbakhsh, and Peter Clark. Self-refine: Iterative refinement with self-feedback. Neural
Information Processing Systems, 2023.
[742] Xinji Mai, Haotian Xu, W. Xing, Weinong Wang, Yingying Zhang, and Wenqiang Zhang. Agent rl
scaling law: Agent rl with spontaneous code execution for mathematical problem solving, arXiv
preprint arXiv:2505.07773, 2025. URL https://arxiv.org/abs/2505.07773v2.
[743] Amjad Yousef Majid, Serge Saaybi, Tomas van Rietbergen, Vincent François-Lavet, R. V. Prasad, and
Chris Verhoeven. Deep reinforcement learning versus evolution strategies: A comparative survey.
IEEE Transactions on Neural Networks and Learning Systems, 2021.
[744] D. Maldonado, Edison Cruz, Jackeline Abad Torres, P. Cruz, and Silvana del Pilar Gamboa Benitez.
Multi-agent systems: A survey about its components, framework and workflow. IEEE Access, 2024.
[745] Zhao Mandi, Shreeya Jain, and Shuran Song. Roco: Dialectic multi-robot collaboration with large
language models. IEEE International Conference on Robotics and Automation, 2023.
[746] Jeremy R. Manning, Sean M. Polyn, G. Baltuch, B. Litt, and M. Kahana. Oscillatory patterns in
temporal lobe reveal context reinstatement during memory search. Proceedings of the National
Academy of Sciences of the United States of America, 2011.
[747] Qiheng Mao, Zemin Liu, Chenghao Liu, Zhuo Li, and Jianling Sun. Advancing graph representa-
tion learning with large language models: A comprehensive survey of techniques, arXiv preprint
arXiv:2402.05952v1, 2024. URL https://arxiv.org/abs/2402.05952v1.
[748] Amin Hosseiny Marani, Ulie Schnaithmann, Youngseo Son, Akil Iyer, Manas Paldhe, and Arushi
Raghuvanshi. Graph integrated language transformers for next action prediction in complex phone
calls. North American Chapter of the Association for Computational Linguistics, 2024.
[749] Sophia Maria. Compass-v2 technical report, arXiv preprint arXiv:2504.15527, 2025. URL https:
//arxiv.org/abs/2504.15527v1.
[750] Viorica Marian and U. Neisser. Language-dependent recall of autobiographical memories. Journal of
experimental psychology. General, 2000.
[751] S. Mariani and Andrea Omicini. Special issue “multi-agent systems”: Editorial. Applied Sciences,
2019.
[752] Vasilije Markovic, Lazar Obradović, L’aszl’o Hajdu, and Jovan Pavlović. Optimizing the interface
between knowledge graphs and llms for complex reasoning, arXiv preprint arXiv:2505.24478, 2025.
URL https://arxiv.org/abs/2505.24478v1.
[753] Sami Marreed, Alon Oved, Avi Yaeli, Segev Shlomov, Ido Levy, Offer Akrabi, Aviad Sela, Asaf Adi,
and Nir Mashkif. Towards enterprise-ready computer using generalist agent. 2025.
[754] Tula Masterman, Sandi Besen, Mason Sawtell, and Alex Chao. The landscape of emerging ai agent
architectures for reasoning, planning, and tool calling: A survey, arXiv preprint arXiv:2404.11584,
2024. URL https://arxiv.org/abs/2404.11584v1.
[755] Nicholas Matsumoto, Jay Moran, Hyunjun Choi, Miguel E. Hernandez, Mythreye Venkatesan, Paul
Wang, and Jason H. Moore. Kragen: a knowledge graph-enhanced rag framework for biomedical
problem solving using large language models. Bioinformatics, 2024.
[756] Ryoga Matsuo, Stefan Uhlich, Arun Venkitaraman, Andrea Bonetti, Chia-Yu Hsieh, Ali Momeni,
Lukas Mauch, Augusto Capone, Eisaku Ohbuchi, and Lorenzo Servadei. Schemato - an llm for
netlist-to-schematic conversion. arXiv preprint, 2024.
[757] Costas Mavromatis, V. Ioannidis, Shen Wang, Da Zheng, Soji Adeshina, Jun Ma, Han Zhao, C. Falout-
sos, and G. Karypis. Train your own gnn teacher: Graph-aware distillation on textual graphs.
ECML/PKDD, 2023.
[758] James L. McClelland, B. McNaughton, and R. O’Reilly. Why there are complementary learning
systems in the hippocampus and neocortex: insights from the successes and failures of connectionist
models of learning and memory. Psychology Review, 1995.
[759] R. Thomas McCoy, Ellie Pavlick, and Tal Linzen. Right for the wrong reasons: Diagnosing syntactic
heuristics in natural language inference. Annual Meeting of the Association for Computational
Linguistics, 2019.
[760] Daniel McDuff, M. Schaekermann, Tao Tu, Anil Palepu, Amy Wang, Jake Garrison, Karan Singhal,
Yash Sharma, Shekoofeh Azizi, Kavita Kulkarni, Le Hou, Yong Cheng, Yun Liu, S. Mahdavi, Sushant
Prakash, Anupam Pathak, Christopher Semturs, Shwetak N. Patel, D. Webster, Ewa Dominowska,
Juraj Gottweis, Joelle Barral, Katherine Chou, G. Corrado, Yossi Matias, Jacob Sunshine, A. Karthike-
salingam, and Vivek Natarajan. Towards accurate differential diagnosis with large language models.
Nature, 2023.
[761] AD McNaughton, G Ramalaxmi, A Kruel, and CR Knutson. . . . Cactus: Chemistry agent connecting
tool-usage to science, arxiv, 2024.
[762] Sushant Mehta, R. Dandekar, R. Dandekar, and S. Panat. Latent multi-head attention for small
language models, arXiv preprint arXiv:2506.09342, 2025. URL https://arxiv.org/abs/2506.
09342v2.
[763] Kai Mei, Zelong Li, Shuyuan Xu, Ruosong Ye, Yingqiang Ge, and Yongfeng Zhang. Aios: Llm agent
operating system, arXiv preprint arXiv:2403.16971, 2024. URL https://arxiv.org/abs/2403.
16971v4.
[764] Lingrui Mei, Shenghua Liu, Yiwei Wang, Baolong Bi, and Xueqi Cheng. Slang: New concept
comprehension of large language models. In Proceedings of the 2024 Conference on Empirical Methods
in Natural Language Processing, page 12558–12575. Association for Computational Linguistics,
2024. doi: 10.18653/v1/2024.emnlp-main.698. URL http://dx.doi.org/10.18653/v1/
2024.emnlp-main.698.
[765] Lingrui Mei, Shenghua Liu, Yiwei Wang, Baolong Bi, Ruibin Yuan, and Xueqi Cheng. Hiddenguard:
Fine-grained safe generation with specialized representation router, arXiv preprint arXiv:2410.02684,
2024. URL https://arxiv.org/abs/2410.02684.
[766] Lingrui Mei, Shenghua Liu, Yiwei Wang, Baolong Bi, Yuyao Ge, Jun Wan, Yurong Wu, and Xueqi
Cheng. a1: Steep test-time scaling law via environment augmented generation, arXiv preprint
arXiv:2504.14597, 2025. URL https://arxiv.org/abs/2504.14597.
[767] Lingrui Mei, Shenghua Liu, Yiwei Wang, Baolong Bi, Jiayi Mao, and Xueqi Cheng. “not aligned” is not
“malicious”: Being careful about hallucinations of large language models’ jailbreak, arXiv preprint
arXiv:2406.11668, 2025. URL https://arxiv.org/abs/2406.11668.
[768] T. Meiser and A. Bröder. Memory for multidimensional source information. Journal of Experimental
Psychology. Learning, Memory and Cognition, 2002.
[769] Gábor Melis, Chris Dyer, and Phil Blunsom. On the state of the art of evaluation in neural language
models. International Conference on Learning Representations, 2017.
[770] Kevin Meng, David Bau, A. Andonian, and Yonatan Belinkov. Locating and editing factual associations
in gpt. Neural Information Processing Systems, 2022.
[771] Yuxian Meng, Shi Zong, Xiaoya Li, Xiaofei Sun, Tianwei Zhang, Fei Wu, and Jiwei Li. Gnn-lm:
Language modeling based on global contexts via gnn. International Conference on Learning Represen-
tations, 2021.
[772] Agnieszka Mensfelt, Kostas Stathis, and Vince Trencsenyi. Towards logically sound natural language
reasoning with logic-enhanced language model agents, arXiv preprint arXiv:2408.16081, 2024. URL
https://arxiv.org/abs/2408.16081v2.
[773] G. M. Mensink and J. Raaijmakers. A model for interference and forgetting. arXiv preprint, 1988.
[774] Thomas Merth, Qichen Fu, Mohammad Rastegari, and Mahyar Najibi. Superposition prompting:
Improving and accelerating retrieval-augmented generation. International Conference on Machine
Learning, 2024.
[775] B. Meskó. Prompt engineering as an important emerging skill for medical professionals: Tutorial.
Journal of Medical Internet Research, 2023.
[776] Yapeng Mi, Zhi Gao, Xiaojian Ma, and Qing Li. Building llm agents by incorporating insights from
computer systems, arXiv preprint arXiv:2504.04485, 2025. URL https://arxiv.org/abs/
2504.04485v1.
[777] G. Mialon, Roberto Dessì, M. Lomeli, Christoforos Nalmpantis, Ramakanth Pasunuru, R. Raileanu,
Baptiste Rozière, Timo Schick, Jane Dwivedi-Yu, Asli Celikyilmaz, Edouard Grave, Yann LeCun, and
Thomas Scialom. Augmented language models: a survey. Trans. Mach. Learn. Res., 2023.
[778] G. Mialon, Clémentine Fourrier, Craig Swift, Thomas Wolf, Yann LeCun, and Thomas Scialom.
Gaia: a benchmark for general ai assistants, arXiv preprint arXiv:2311.12983, 2023. URL https:
//arxiv.org/abs/2311.12983v1.
[779] Xupeng Miao, Gabriele Oliaro, Zhihao Zhang, Xinhao Cheng, Hongyi Jin, Tianqi Chen, and Zhihao
Jia. Towards efficient generative large language model serving: A survey from algorithms to systems,
arXiv preprint arXiv:2312.15234, 2023. URL https://arxiv.org/abs/2312.15234v1.
[780] Jacob Miller, Guillaume Rabusseau, and John Terilla. Tensor networks for language modeling. arXiv
preprint, 2020.
[781] Xing ming Guo, Darioush Keivan, U. Syed, Lianhui Qin, Huan Zhang, G. Dullerud, Peter J. Seiler,
and Bin Hu. Controlagent: Automating control system design via novel integration of llm agents
and domain expertise, arXiv preprint arXiv:2410.19811, 2024. URL https://arxiv.org/abs/
2410.19811v1.
[782] Soroush Mirjalili, Patrick S. Powell, Jonathan Strunk, Taylor A James, and Audrey Duarte. Context
memory encoding and retrieval temporal dynamics are modulated by attention across the adult
lifespan. eNeuro, 2021.
[783] Ishan Misra and L. Maaten. Self-supervised learning of pretext-invariant representations. Computer
Vision and Pattern Recognition, 2019.
[784] Ali Modarressi, Ayyoob Imani, Mohsen Fayyaz, and Hinrich Schütze. Ret-llm: Towards a general
read-write memory for large language models, arXiv preprint arXiv:2305.14322, 2024. URL https:
//arxiv.org/abs/2305.14322.
[785] Ali Modarressi, Abdullatif Köksal, Ayyoob Imani, Mohsen Fayyaz, and Hinrich Schutze. Memllm:
Finetuning llms to use an explicit read-write memory. Trans. Mach. Learn. Res., 2024.
[786] Behnam Mohammadi. Pel, a programming language for orchestrating ai agents, arXiv preprint
arXiv:2505.13453, 2025. URL https://arxiv.org/abs/2505.13453v2.
[787] Amirkeivan Mohtashami and Martin Jaggi. Landmark attention: Random-access infinite context
length for transformers. Neural Information Processing Systems, 2023.
[788] Fedor Moiseev, Zhe Dong, Enrique Alfonseca, and Martin Jaggi. Skill: Structured knowledge infusion
for large language models. North American Chapter of the Association for Computational Linguistics,
2022.
[789] Dimitri Coelho Mollo and Raphael Milliere. The vector grounding problem, arXiv preprint
arXiv:2304.01481, 2023. URL https://arxiv.org/abs/2304.01481v2.
[790] Nieves Montes, N. Osman, and C. Sierra. Combining theory of mind and abduction for cooperation
under imperfect information. European Workshop on Multi-Agent Systems, 2022.
[791] Suhong Moon, Siddharth Jha, Lutfi Eren Erdogan, Sehoon Kim, Woosang Lim, Kurt Keutzer, and
A. Gholami. Efficient and scalable estimation of tool representations in vector space, arXiv preprint
arXiv:2409.02141, 2024. URL https://arxiv.org/abs/2409.02141v1.
[792] Shinsuke Mori. A stochastic parser based on an slm with arboreal context trees. International
Conference on Computational Linguistics, 2002.
[793] Meredith Ringel Morris. Prompting considered harmful. Communications of the ACM, 2024.
[794] Marius Mosbach, Tiago Pimentel, Shauli Ravfogel, D. Klakow, and Yanai Elazar. Few-shot fine-tuning
vs. in-context learning: A fair comparison and evaluation. Annual Meeting of the Association for
Computational Linguistics, 2023.
[795] Sajad Mousavi, Ricardo Luna Guti’errez, Desik Rengarajan, Vineet Gundecha, Ashwin Ramesh Babu,
Avisek Naug, Antonio Guillen-Perez, and S. Sarkar. N-critics: Self-refinement of large language
models with ensemble of critics. arXiv preprint, 2023.
[796] Manisha Mukherjee, Sungchul Kim, Xiang Chen, Dan Luo, Tong Yu, and Tung Mai. From documents
to dialogue: Building kg-rag enhanced ai assistants, arXiv preprint arXiv:2502.15237, 2025. URL
https://arxiv.org/abs/2502.15237v1.
[797] Tergel Munkhbat, Namgyu Ho, Seohyun Kim, Yongjin Yang, Yujin Kim, and Se young Yun. Self-
training elicits concise reasoning in large language models, arXiv preprint arXiv:2502.20122, 2025.URL https://arxiv.org/abs/2502.20122v3.
[798] Tsendsuren Munkhdalai, Manaal Faruqui, and Siddharth Gopal. Leave no context behind: Efficient infinite context transformers with infini-attention, arXiv preprint arXiv:2404.07143, 2024. URL https://arxiv.org/abs/2404.07143v2.
[799] Eliya Nachmani, Alon Levkovitch, Julián Salazar, Chulayutsh Asawaroengchai, Soroosh Mariooryad,
R. Skerry-Ryan, and Michelle Tadmor Ramanovich. Spoken question answering and speech con-
tinuation using spectrogram-powered llm. International Conference on Learning Representations,
2023.
[800] L. Nadel, Jessica D. Payne, and W. J. Jacobs. The relationship between episodic memory and context:
clues from memory errors made while under stress. Physiological Research, 2002.
[801] Reiichiro Nakano, Jacob Hilton, Suchir Balaji, Jeff Wu, Long Ouyang, Christina Kim, Christopher
Hesse, Shantanu Jain, Vineet Kosaraju, William Saunders, Xu Jiang, Karl Cobbe, Tyna Eloundou,
Gretchen Krueger, Kevin Button, Matthew Knight, Benjamin Chess, and John Schulman. Webgpt:
Browser-assisted question-answering with human feedback, arXiv preprint arXiv:2112.09332, 2022.
URL https://arxiv.org/abs/2112.09332.
[802] Koichi Namekata, Amirmojtaba Sabour, Sanja Fidler, and Seung Wook Kim. Emerdiff: Emerging
pixel-level semantic knowledge in diffusion models, arXiv preprint arXiv:2401.11739, 2024. URL
https://arxiv.org/abs/2401.11739.
[803] Sundaraparipurnan Narayanan and Sandeep Vishwakarma. Guard-d-llm: An llm-based risk assess-
ment engine for the downstream uses of llms. arXiv preprint, 2024.
[804] Usman Naseem, Surendrabikram Thapa, Qi Zhang, Liang Hu, Anum Masood, and Mehwish Nasim.
Reducing knowledge noise for improved semantic analysis in biomedical natural language processing
applications. Clinical Natural Language Processing Workshop, 2023.
[805] Deepak Nathani, David Wang, Liangming Pan, and W. Wang. Maf: Multi-aspect feedback for
improving reasoning in large language models. Conference on Empirical Methods in Natural Language
Processing, 2023.
[806] Aashutosh Nema, Samaksh Gulati, Evangelos Giakoumakis, and Bipana Thapaliya. Modp: Multi
objective directional prompting, arXiv preprint arXiv:2504.18722, 2025. URL https://arxiv.
org/abs/2504.18722v1.
[807] Christian D. Newman, Anthony S Peruma, and Reem S. Alsuhaibani. Modeling the relationship
between identifier name and behavior. IEEE International Conference on Software Maintenance and
Evolution, 2019.
[808] M. Nieznański, Michał Obidziński, Emilia Zyskowska, and Daria Niedziałkowska. Executive resources
and item-context binding: Exploring the influence of concurrent inhibition, updating, and shifting
tasks on context memory. Advances in Cognitive Psychology, 2015.
[809] M. Nieznański, Michał Obidziński, and Daria Ford. Does context recollection depend on the base-rate
of contextual features? Cognitive Processing, 2023.
[810] C. Nourani and P. Eklund. Concepts ontology algebras and role descriptions. Conference on Computer
Science and Information Systems, 2017.
[811] Felix Ocker, Daniel Tanneberg, Julian Eggert, and Michael Gienger. Tulip agent - enabling llm-based
agents to solve tasks using large tool libraries. arXiv preprint, 2024.
[812] Felix Ocker, J. Deigmöller, Pavel Smirnov, and Julian Eggert. A grounded memory system for smart
personal assistants, arXiv preprint arXiv:2505.06328, 2025. URL https://arxiv.org/abs/
2505.06328v1.
[813] OpenAI. Computer-using agent, 2025. URL https://openai.com/index/
computer-using-agent/. OpenAI Technical Report.
[814] OpenAI. Swarm: Educational framework exploring ergonomic, lightweight multi-agent orchestration.
https://github.com/openai/swarm, 2025. [Online; accessed 17-July-2025].
[815] Jonas Oppenlaender. Dangermaps: Personalized safety advice for travel in urban environments
using a retrieval-augmented language model, arXiv preprint arXiv:2503.14103, 2025. URL https:
//arxiv.org/abs/2503.14103v3.
[816] A. Orhan. Recognition, recall, and retention of few-shot memories in large language models, arXiv
preprint arXiv:2303.17557, 2023. URL https://arxiv.org/abs/2303.17557v1.
[817] Gustavo Ortiz-Hernández, Alejandro Guerra-Hernández, J. Hübner, and W. A. Luna-Ramírez. Mod-
ularization in belief-desire-intention agent programming and artifact-based environments. PeerJ
Computer Science, 2022.
[818] Wendkûuni C. Ouédraogo, A. Kaboré, Haoye Tian, Yewei Song, Anil Koyuncu, Jacques Klein, David
Lo, and Tegawend’e F. Bissyand’e. Large-scale, independent and comprehensive study of the power
of llms for test case generation. arXiv preprint, 2024.
[819] Charles Packer, Vivian Fang, Shishir G. Patil, Kevin Lin, Sarah Wooders, and Joseph Gonzalez.
Memgpt: Towards llms as operating systems, arXiv preprint arXiv:2310.08560, 2023. URL https:
//arxiv.org/abs/2310.08560v2.
[820] Charles Packer, Sarah Wooders, Kevin Lin, Vivian Fang, Shishir G. Patil, Ion Stoica, and Joseph E.
Gonzalez. Memgpt: Towards llms as operating systems, arXiv preprint arXiv:2310.08560, 2024.
URL https://arxiv.org/abs/2310.08560.
[821] Constantin-Valentin Pal, F. Leon, M. Paprzycki, and M. Ganzha. A review of platforms for the
development of agent systems. Inf., 2020.
[822] Qianjun Pan, Wenkai Ji, Yuyang Ding, Junsong Li, Shilian Chen, Junyi Wang, Jie Zhou, Qin Chen, Min
Zhang, Yulan Wu, and Liang He. A survey of slow thinking-based reasoning llms using reinforced
learning and inference-time scaling law, arXiv preprint arXiv:2505.02665, 2025. URL https:
//arxiv.org/abs/2505.02665v2.
[823] Shirui Pan, Linhao Luo, Yufei Wang, Chen Chen, Jiapu Wang, and Xindong Wu. Unifying large
language models and knowledge graphs: A roadmap. IEEE Transactions on Knowledge and Data
Engineering, 2023.
[824] Xu Pan, Ely Hahami, Zechen Zhang, and H. Sompolinsky. Memorization and knowledge injection
in gated llms, arXiv preprint arXiv:2504.21239, 2025. URL https://arxiv.org/abs/2504.
21239v1.
[825] Bo Pang, Hanze Dong, Jiacheng Xu, Silvio Savarese, Yingbo Zhou, and Caiming Xiong. Bolt: Bootstrap
long chain-of-thought in language models without distillation, arXiv preprint arXiv:2502.03860,
2025. URL https://arxiv.org/abs/2502.03860v1.
[826] Jianhui Pang, Fanghua Ye, Derek F. Wong, and Longyue Wang. Anchor-based large language models.
Annual Meeting of the Association for Computational Linguistics, 2024.
[827] Bhargavi Paranjape, Scott M. Lundberg, Sameer Singh, Hannaneh Hajishirzi, Luke Zettlemoyer, and
Marco Tulio Ribeiro. Art: Automatic multi-step reasoning and tool-use for large language models,
arXiv preprint arXiv:2303.09014, 2023. URL https://arxiv.org/abs/2303.09014v1.
[828] A Parisi, Y Zhao, and N Fiedel. Talm: Tool augmented language models. 2022. URL https:
//arxiv.org/abs/2205.12255.
[829] Dongju Park and Chang Wook Ahn. Self-supervised contextual data augmentation for natural
language processing. Symmetry, 2019.
[830] J. Park, Lindsay Popowski, Carrie J. Cai, M. Morris, Percy Liang, and Michael S. Bernstein. Social
simulacra: Creating populated prototypes for social computing systems. ACM Symposium on User
Interface Software and Technology, 2022.
[831] J. Park, Joseph C. O’Brien, Carrie J. Cai, M. Morris, Percy Liang, and Michael S. Bernstein. Generative
agents: Interactive simulacra of human behavior. ACM Symposium on User Interface Software and
Technology, 2023.
[832] Soya Park, J. Zamfirescu-Pereira, and Chinmay Kulkarni. Model behavior specification by leveraging
llm self-playing and self-improving, arXiv preprint arXiv:2503.03967, 2025. URL https://arxiv.
org/abs/2503.03967v1.
[833] Rajvardhan Patil and Venkat Gudivada. A review of current trends, techniques, and challenges in
large language models (llms). Applied Sciences, 2024.
[834] Shishir G. Patil, Tianjun Zhang, Xin Wang, and Joseph E. Gonzalez. Gorilla: Large language model
connected with massive apis, arXiv preprint arXiv:2305.15334, 2023. URL https://arxiv.org/
abs/2305.15334.
[835] Shishir G. Patil, Huanzhi Mao, Charlie Cheng-Jie Ji, Fanjia Yan, Vishnu Suresh, Ion Stoica, and
Joseph E. Gonzalez. The berkeley function calling leaderboard (bfcl): From tool use to agentic
evaluation of large language models. In Forty-second International Conference on Machine Learning,
2025.
[836] Shuva Paul, Farhad Alemi, and Richard Macwan. Llm-assisted proactive threat intelligence for
automated reasoning, arXiv preprint arXiv:2504.00428, 2025. URL https://arxiv.org/abs/
2504.00428v1.
[837] Saurav Pawar, S. Tonmoy, S. M. M. Zaman, Vinija Jain, Aman Chadha, and Amitava Das. The what,
why, and how of context length extension techniques in large language models - a detailed survey.
arXiv preprint, 2024.
[838] Boci Peng, Yun Zhu, Yongchao Liu, Xiaohe Bo, Haizhou Shi, Chuntao Hong, Yan Zhang, and
Siliang Tang. Graph retrieval-augmented generation: A survey. ArXiv, abs/2408.08921, 2024. URL
https://api.semanticscholar.org/CorpusID:271903170.
[839] Bowen Peng, Jeffrey Quesnelle, Honglu Fan, and Enrico Shippole. Yarn: Efficient context window
extension of large language models. International Conference on Learning Representations, 2023.
[840] Hao Peng, Tianyu Gao, Xu Han, Yankai Lin, Peng Li, Zhiyuan Liu, Maosong Sun, and Jie Zhou.
Learning from context or names? an empirical study on neural relation extraction. Conference on
Empirical Methods in Natural Language Processing, 2020.
[841] Ji-Lun Peng, Sijia Cheng, Egil Diau, Yung-Yu Shih, Po-Heng Chen, Yen-Ting Lin, and Yun-Nung
Chen. A survey of useful llm evaluation, arXiv preprint arXiv:2406.00936, 2024. URL https:
//arxiv.org/abs/2406.00936v1.
[842] Bryan Perozzi, Bahare Fatemi, Dustin Zelle, Anton Tsitsulin, Mehran Kazemi, Rami Al-Rfou, and
Jonathan J. Halcrow. Let your graph do the talking: Encoding structured data for llms, arXiv preprint
arXiv:2402.05862, 2024. URL https://arxiv.org/abs/2402.05862v1.
[843] E. Pesce and G. Montana. Improving coordination in small-scale multi-agent deep reinforcement
learning through memory-driven communication. Machine-mediated learning, 2019.
[844] F. Petroni, Patrick Lewis, Aleksandra Piktus, Tim Rocktäschel, Yuxiang Wu, Alexander H. Miller, and
Sebastian Riedel. How context affects language models’ factual predictions. Conference on Automated
Knowledge Base Construction, 2020.
[845] Yue Pi, Wang Zhang, Yong Zhang, Hairong Huang, Baoquan Rao, Yulong Ding, and Shuanghua
Yang. Applications of multi-agent deep reinforcement learning communication in network manage-
ment: A survey, arXiv preprint arXiv:2407.17030, 2024. URL https://arxiv.org/abs/2407.
17030v1.
[846] Nancirose Piazza and Vahid Behzadan. A theory of mind approach as test-time mitigation against
emergent adversarial communication. Adaptive Agents and Multi-Agent Systems, 2023.
[847] Mathis Pink, Vy A. Vo, Qinyuan Wu, Jianing Mu, Javier S. Turek, Uri Hasson, Kenneth A. Norman,
Sebastian Michelmann, Alexander Huth, and Mariya Toneva. Assessing episodic memory in llms
with sequence order recall tasks, arXiv preprint arXiv:2410.08133, 2024. URL https://arxiv.
org/abs/2410.08133v1.
[848] Fahmida Liza Piya and Rahmatollah Beheshti. Advancing feature extraction in healthcare through
the integration of knowledge graphs and large language models. AAAI Conference on Artificial
Intelligence, 2025.
[849] A. Plaat, M. V. Duijn, N. V. Stein, Mike Preuss, P. V. D. Putten, and K. Batenburg. Agentic large
language models, a survey, arXiv preprint arXiv:2503.23037, 2025. URL https://arxiv.org/
abs/2503.23037v2.
[850] Moritz Plenz and Anette Frank. Graph language models. Annual Meeting of the Association for
Computational Linguistics, 2024.
[851] Sean M. Polyn, K. Norman, and M. Kahana. A context maintenance and retrieval model of organiza-
tional processes in free recall. Psychology Review, 2009.
[852] Liam Pond and Ichiro Fujinaga. Teaching llms music theory with in-context learning and chain-
of-thought prompting: Pedagogical strategies for machines. International Conference on Computer
Supported Education, 2025.
[853] V Porcu. The role of memory in llms: Persistent context for smarter conversations. Int. J. Sci. Res.
Manag.(IJSRM), 12:1673–1691, 2024.
[854] Ofir Press, Noah A. Smith, and M. Lewis. Train short, test long: Attention with linear biases enables
input length extrapolation. International Conference on Learning Representations, 2021.
[855] Xavier Puig, K. Ra, Marko Boben, Jiaman Li, Tingwu Wang, S. Fidler, and A. Torralba. Virtualhome:
Simulating household activities via programs. 2018 IEEE/CVF Conference on Computer Vision and
Pattern Recognition, 2018.
[856] Pranav Putta, Edmund Mills, Naman Garg, S. Motwani, Chelsea Finn, Divyansh Garg, and Rafael
Rafailov. Agent q: Advanced reasoning and learning for autonomous ai agents, arXiv preprint
arXiv:2408.07199, 2024. URL https://arxiv.org/abs/2408.07199v1.
[857] S. Qasim, Hassan Mahmood, and F. Shafait. Rethinking table recognition using graph neural networks.
IEEE International Conference on Document Analysis and Recognition, 2019.
[858] Peng Qi, Haejun Lee, OghenetegiriTGSido, and Christopher D. Manning. Answering open-domain
questions of varying reasoning steps from text. Conference on Empirical Methods in Natural Language
Processing, 2020.
[859] Yong Qi, Gabriel Kyebambo, Siyuan Xie, Wei Shen, Shenghui Wang, Bitao Xie, Bin He, Zhipeng
Wang, and Shuo Jiang. Safety control of service robots with llms and embodied knowledge graphs,
arXiv preprint arXiv:2405.17846, 2024. URL https://arxiv.org/abs/2405.17846v1.
[860] Zehan Qi, Xiao Liu, Iat Long Iong, Hanyu Lai, Xueqiao Sun, Xinyue Yang, Jiadai Sun, Yu Yang,
Shuntian Yao, Tianjie Zhang, Wei Xu, Jie Tang, and Yuxiao Dong. Webrl: Training llm web agents
via self-evolving online curriculum reinforcement learning, arXiv preprint arXiv:2411.02337, 2024.
URL https://arxiv.org/abs/2411.02337v3.
[861] Chen Qian, Wei Liu, Hongzhang Liu, Nuo Chen, Yufan Dang, Jiahao Li, Cheng Yang, Weize Chen,
Yusheng Su, Xin Cong, Juyuan Xu, Dahai Li, Zhiyuan Liu, and Maosong Sun. Chatdev: Communicative
agents for software development, arXiv preprint arXiv:2307.07924, 2024. URL https://arxiv.
org/abs/2307.07924.
[862] Cheng Qian, Chi Han, Y. Fung, Yujia Qin, Zhiyuan Liu, and Heng Ji. Creator: Tool creation for
disentangling abstract and concrete reasoning of large language models. Conference on Empirical
Methods in Natural Language Processing, 2023.
[863] Cheng Qian, Jiahao Li, Yufan Dang, Wei Liu, Yifei Wang, Zihao Xie, Weize Chen, Cheng Yang,
Yingli Zhang, Zhiyuan Liu, and Maosong Sun. Iterative experience refinement of software-
developing agents, arXiv preprint arXiv:2405.04219, 2024. URL https://arxiv.org/abs/
2405.04219v1.
[864] Cheng Qian, Emre Can Acikgoz, Qi He, Hongru Wang, Xiusi Chen, Dilek Hakkani-Tur, Gokhan Tur,
and Heng Ji. Toolrl: Reward is all tool learning needs, arXiv preprint arXiv:2504.13958, 2025. URL
https://arxiv.org/abs/2504.13958v1.
[865] Hongjin Qian, Zheng Liu, Peitian Zhang, Zhicheng Dou, and Defu Lian. Boosting long-context
management via query-guided activation refilling, arXiv preprint arXiv:2412.12486, 2024. URL
https://arxiv.org/abs/2412.12486v3.
[866] Hongjin Qian, Zheng Liu, Peitian Zhang, Kelong Mao, Defu Lian, Zhicheng Dou, and Tiejun Huang.
Memorag: Boosting long context processing with global memory-enhanced retrieval augmentation,
arXiv preprint arXiv:2409.05591, 2025. URL https://arxiv.org/abs/2409.05591.
[867] Kangan Qian, Sicong Jiang, Yang Zhong, Ziang Luo, Zilin Huang, Tianze Zhu, Kun Jiang, Mengmeng
Yang, Zheng Fu, Jinyu Miao, Yining Shi, He Zhe Lim, Li Liu, Tianbao Zhou, Hongyi Wang, Huang
Yu, Yifei Hu, Guang Li, Guangyao Chen, Hao Ye, Lijun Sun, and Diange Yang. Agentthink: A
unified framework for tool-augmented chain-of-thought reasoning in vision-language models for
autonomous driving, arXiv preprint arXiv:2505.15298, 2025. URL https://arxiv.org/abs/
2505.15298v3.
[868] Changze Qiao and Mingming Lu. Efficiently enhancing general agents with hierarchical-categorical
memory, arXiv preprint arXiv:2505.22006, 2025. URL https://arxiv.org/abs/2505.
22006v1.
[869] S Qiao, H Gui, C Lv, Q Jia, H Chen, and N Zhang. Making language models better tool learners with
execution feedback. 2023. URL https://arxiv.org/abs/2305.13068.
[870] Binjie Qin, Haohao Mao, Ruipeng Zhang, Y. Zhu, Song Ding, and Xu Chen. Working memory
inspired hierarchical video decomposition with transformative representations, arXiv preprint
arXiv:2204.10105, 2022. URL https://arxiv.org/abs/2204.10105v3.
[871] Bowen Qin, Binyuan Hui, Lihan Wang, Min Yang, Jinyang Li, Binhua Li, Ruiying Geng, Rongyu Cao,
Jian Sun, Luo Si, Fei Huang, and Yongbin Li. A survey on text-to-sql parsing: Concepts, methods,
and future directions, arXiv preprint arXiv:2208.13629, 2022. URL https://arxiv.org/abs/
2208.13629v1.
[872] Libo Qin, Qiguang Chen, Fuxuan Wei, Shijue Huang, and Wanxiang Che. Cross-lingual prompting:
Improving zero-shot chain-of-thought reasoning across languages, arXiv preprint arXiv:2310.14799,
2023. URL https://arxiv.org/abs/2310.14799.
[873] Libo Qin, Fuxuan Wei, Qiguang Chen, Jingxuan Zhou, Shijue Huang, Jiasheng Si, Wenpeng Lu,
and Wanxiang Che. Croprompt: Cross-task interactive prompting for zero-shot spoken language
understanding, arXiv preprint arXiv:2406.10505, 2024. URL https://arxiv.org/abs/2406.
10505.
[874] Yujia Qin, Shengding Hu, Yankai Lin, Weize Chen, Ning Ding, Ganqu Cui, Zheni Zeng, Yufei Huang,
Chaojun Xiao, Chi Han, Y. Fung, Yusheng Su, Huadong Wang, Cheng Qian, Runchu Tian, Kunlun
Zhu, Shi Liang, Xingyu Shen, Bokai Xu, Zhen Zhang, Yining Ye, Bo Li, Ziwei Tang, Jing Yi, Yu Zhu,
Zhenning Dai, Lan Yan, Xin Cong, Ya-Ting Lu, Weilin Zhao, Yuxiang Huang, Junxi Yan, Xu Han,
Xian Sun, Dahai Li, Jason Phang, Cheng Yang, Tongshuang Wu, Heng Ji, Zhiyuan Liu, and Maosong
Sun. Tool learning with foundation models. ACM Computing Surveys, 2023.
[875] Yujia Qin, Shi Liang, Yining Ye, Kunlun Zhu, Lan Yan, Ya-Ting Lu, Yankai Lin, Xin Cong, Xiangru Tang,
Bill Qian, Sihan Zhao, Runchu Tian, Ruobing Xie, Jie Zhou, Marc H. Gerstein, Dahai Li, Zhiyuan Liu,
and Maosong Sun. Toolllm: Facilitating large language models to master 16000+ real-world apis.
International Conference on Learning Representations, 2023.
[876] Zhen Qin and Yiran Zhong. Accelerating toeplitz neural network with constant-time inference
complexity. Conference on Empirical Methods in Natural Language Processing, 2023.
[877] Zhen Qin, Xiaodong Han, Weixuan Sun, Bowen He, Dong Li, Dongxu Li, Yuchao Dai, Lingpeng
Kong, and Yiran Zhong. Toeplitz neural network for sequence modeling. International Conference on
Learning Representations, 2023.
[878] Zhen Qin, Weigao Sun, Dong Li, Xuyang Shen, Weixuan Sun, and Yiran Zhong. Lightning attention-2:
A free lunch for handling unlimited sequence lengths in large language models. arXiv preprint, 2024.
[879] Jiahao Qiu, Xinzhe Juan, Yiming Wang, Ling Yang, Xuan Qi, Tongcheng Zhang, Jiacheng Guo, Yifu
Lu, Zixin Yao, Hongru Wang, Shilong Liu, Xun Jiang, Liu Leqi, and Mengdi Wang. Agentdistill:
Training-free agent distillation with generalizable mcp boxes, arXiv preprint arXiv:2506.14728,
2025. URL https://arxiv.org/abs/2506.14728v1.
[880] Jiahao Qiu, Fulian Xiao, Yiming Wang, Yuchen Mao, Yijia Chen, Xinzhe Juan, Siran Wang, Xuan
Qi, Tongcheng Zhang, Zixin Yao, Jiacheng Guo, Yifu Lu, Charles Argon, Jundi Cui, Daixin Chen,
Junran Zhou, Shuyao Zhou, Zhanpeng Zhou, Ling Yang, Shilong Liu, Hongru Wang, Kaixuan
Huang, Xun Jiang, Yuming Cao, Yue Chen, Yunfei Chen, Zhengyi Chen, Ruowei Dai, Mengqiu
Deng, Jiye Fu, Yu Gu, Zijie Guan, Zirui Huang, Xiaoyan Ji, Yumeng Jiang, Delong Kong, Haolong
Li, Jiaqi Li, Ruipeng Li, Tianze Li, Zhuo-Yang Li, Haixia Lian, Meng Lin, Xudong Liu, Jiayi Lu,
Jinghan Lu, Wanyu Luo, Ziyue Luo, Zihao Pu, Zhi Qiao, Rui-Fang Ren, Liang Wan, Ruixiang Wang,
Tianhui Wang, Yang Wang, Zeyu Wang, Zihua Wang, Yujia Wu, Zhaoyi Wu, Hao Xin, Weiao Xing,
Ruojun Xiong, Weijie Xu, Yao Shu, Xiao Yao, Xiaorui Yang, Yuchen Yang, Nan Yi, Jiadong Yu, Yang
Yu, Huiting Zeng, Danni Zhang, Yunjie Zhang, Zhaoyu Zhang, Zhiheng Zhang, Xiaofeng Zheng,
Peirong Zhou, Li-Ying Zhong, Xiaoyin Zong, Ying Zhao, Zhen Chen, Lin Ding, Xiaoyu Gao, Bingbing
Gong, Yichao Li, Yang Liao, Guang Ma, Tianyuan Ma, Xinrui Sun, Tianyi Wang, Han Xia, Ruobing
Xian, Gen Ye, Tengfei Yu, Wentao Zhang, Yuxi Wang, Xi Gao, and Mengdi Wang. On path to
multimodal historical reasoning: Histbench and histagent, arXiv preprint arXiv:2505.20246, 2025.
URL https://arxiv.org/abs/2505.20246v3.
[881] Ruidi Qiu, Grace Li Zhang, Rolf Drechsler, Ulf Schlichtmann, and Bing Li. Autobench: Automatic
testbench generation and evaluation using llms for hdl design. Workshop on Machine Learning for
CAD, 2024.
[882] Changle Qu, Sunhao Dai, Xiaochi Wei, Hengyi Cai, Shuaiqiang Wang, Dawei Yin, Jun Xu, and Jirong
Wen. Tool learning with large language models: A survey. Frontiers Comput. Sci., 2024.
[883] Xiaoye Qu, Yafu Li, Zhao yu Su, Weigao Sun, Jianhao Yan, Dongrui Liu, Ganqu Cui, Daizong
Liu, Shuxian Liang, Junxian He, Peng Li, Wei Wei, Jing Shao, Chaochao Lu, Yue Zhang, Xian-
Sheng Hua, Bowen Zhou, and Yu Cheng. A survey of efficient reasoning for large reasoning
models: Language, multimodality, and beyond, arXiv preprint arXiv:2503.21614, 2025. URL
https://arxiv.org/abs/2503.21614v1.
[884] Victor Quintanar-Zilinskas. A neuromimetic approach to the serial acquisition, long-term storage,
and selective utilization of overlapping memory engrams. bioRxiv, 2019.
[885] Stephan Raaijmakers, Roos Bakker, Anita Cremers, Roy de Kleijn, Tom Kouwenhoven, and Tessa
Verhoef. Memory-augmented generative adversarial transformers, arXiv preprint arXiv:2402.19218,
2024. URL https://arxiv.org/abs/2402.19218.
[886] Ella Rabinovich and Ateret Anaby-Tavor. On the robustness of agentic function calling. Proceedings
of the 5th Workshop on Trustworthy NLP (TrustNLP 2025), 2025.
[887] Neil C. Rabinowitz, Frank Perbet, H. F. Song, Chiyuan Zhang, S. Eslami, and M. Botvinick. Machine
theory of mind. International Conference on Machine Learning, 2018.
[888] Zackary Rackauckas. Rag-fusion: a new take on retrieval-augmented generation. International
Journal on Natural Language Computing, 2024.
[889] Alec Radford, Jong Wook Kim, Chris Hallacy, A. Ramesh, Gabriel Goh, Sandhini Agarwal, Girish
Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and I. Sutskever. Learning
transferable visual models from natural language supervision. International Conference on Machine
Learning, 2021.
[890] Colin Raffel, Noam M. Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi
Zhou, Wei Li, and Peter J. Liu. Exploring the limits of transfer learning with a unified text-to-text
transformer. Journal of machine learning research, 2019.
[891] Mohaimenul Azam Khan Raiaan, Md. Saddam Hossain Mukta, Kaniz Fatema, Nur Mohammad
Fahad, S. Sakib, Most. Marufatul Jannat Mim, Jubaer Ahmad, Mohammed Eunus Ali, and Sami
Azam. A review on large language models: Architectures, applications, taxonomies, open issues and
challenges. IEEE Access, 2024.
[892] Keshav Ramji, Young-Suk Lee, R. Astudillo, M. Sultan, Tahira Naseem, Asim Munawar, Radu Florian,
and S. Roukos. Self-refinement of language models from external proxy metrics feedback, arXiv
preprint arXiv:2403.00827, 2024. URL https://arxiv.org/abs/2403.00827v1.
[893] Sumedh Rasal. An artificial neuron for enhanced problem solving in large language models, arXiv
preprint arXiv:2404.14222, 2024. URL https://arxiv.org/abs/2404.14222v1.
[894] Shaina Raza, Ranjan Sapkota, Manoj Karkee, and Christos Emmanouilidis. Trism for agentic ai:
A review of trust, risk, and security management in llm-based agentic multi-agent systems, arXiv
preprint arXiv:2506.04133, 2025. URL https://arxiv.org/abs/2506.04133v2.
[895] Jing Ren and Feng Xia. Brain-inspired artificial intelligence: A comprehensive review, arXiv preprint
arXiv:2408.14811, 2024. URL https://arxiv.org/abs/2408.14811v1.
[896] Shuo Ren, Pu Jian, Zhenjiang Ren, Chunlin Leng, Can Xie, and Jiajun Zhang. Towards scientific
intelligence: A survey of llm-based scientific agents, arXiv preprint arXiv:2503.24047, 2025. URL
https://arxiv.org/abs/2503.24047v2.
[897] Xubin Ren, Jiabin Tang, Dawei Yin, Nitesh V. Chawla, and Chao Huang. A survey of large language
models for graphs. Knowledge Discovery and Data Mining, 2024.
[898] Yi Ren, Shangmin Guo, Linlu Qiu, Bailin Wang, and Danica J. Sutherland. Bias amplification in
language model evolution: An iterated learning perspective. Neural Information Processing Systems,
2024.
[899] Alireza Rezazadeh, Zichao Li, Wei Wei, and Yujia Bao. From isolated conversations to hierarchical
schemas: Dynamic tree memory representation for llms. International Conference on Learning
Representations, 2024.
[900] Marco Tulio Ribeiro, Carlos Guestrin, and Sameer Singh. Are red roses red? evaluating consistency
of question-answering models. Annual Meeting of the Association for Computational Linguistics, 2019.
[901] Yara Rizk, Abhishek Bhandwalder, S. Boag, Tathagata Chakraborti, Vatche Isahagian, Y. Khazaeni,Falk Pollock, and Merve Unuvar. A unified conversational assistant framework for business processautomation, arXiv preprint arXiv:2001.03543, 2020. URL https://arxiv.org/abs/2001.03543v1.
[902] Yara Rizk, Vatche Isahagian, S. Boag, Y. Khazaeni, Merve Unuvar, Vinod Muthusamy, and Rania Y.Khalaf. A conversational digital assistant for intelligent process automation. International Conference on Business Process Management, 2020.
[903] S. Rizvi, Nazreen Pallikkavaliyaveetil, David Zhang, Zhuoyang Lyu, Nhi Nguyen, Haoran Lyu,
B. Christensen, J. O. Caro, Antonio H. O. Fonseca, E. Zappala, Maryam Bagherian, Christopher
Averill, C. Abdallah, Amin Karbasi, Rex Ying, M. Brbic, R. M. Dhodapkar, and David van Dijk.
Fimp: Foundation model-informed message passing for graph neural networks, arXiv preprint
arXiv:2210.09475v5, 2022. URL https://arxiv.org/abs/2210.09475v5.
[904] Joshua Robinson, Christopher Rytting, and D. Wingate. Leveraging large language models for
multiple choice question answering. International Conference on Learning Representations, 2022.
[905] Juri Di Rocco, D. D. Ruscio, Claudio Di Sipio, P. T. Nguyen, and Riccardo Rubei. On the use of large
language models in model-driven engineering. Journal of Software and Systems Modeling, 2024.
[906] Juan David Salazar Rodriguez, Sam Conrad Joyce, and Julfendi Julfendi. Using customized gpt to develop prompting proficiency in architectural ai-generated images, arXiv preprint arXiv:2504.13948, 2025. URL https://arxiv.org/abs/2504.13948v2.
[907] Albert Roethel, M. Ganzha, and Anna Wr’oblewska. Enriching language models with graph-based
context information to better understand textual data. Electronics, 2023.
[908] Reudismam Rolim, Gustavo Soares, Rohit Gheyi, and Loris D’antoni. Learning quick fixes from code
repositories. Brazilian Symposium on Software Engineering, 2018.
[909] Robin Rombach, A. Blattmann, Dominik Lorenz, Patrick Esser, and B. Ommer. High-resolution image synthesis with latent diffusion models. Computer Vision and Pattern Recognition, 2021.
[910] Hayley Ross, A. Mahabaleshwarkar, and Yoshi Suhara. When2call: When (not) to call tools. North American Chapter of the Association for Computational Linguistics, 2025.
[911] J. Rosser and Jakob N. Foerster. Agentbreeder: Mitigating the ai safety impact of multi-agent scaffolds, arXiv preprint arXiv:2502.00757, 2025. URL https://arxiv.org/abs/2502.00757v3.
[912] Federico Rossi, Saptarshi Bandyopadhyay, Michael T. Wolf, and M. Pavone. Review of multi-agent algorithms for collective behavior: a structural taxonomy, arXiv preprint arXiv:1803.05464, 2018. URL https://arxiv.org/abs/1803.05464v1.
[913] Federico Rossi, Saptarshi Bandyopadhyay, Michael T. Wolf, and M. Pavone. Multi-agent algorithms for collective behavior: A structural and application-focused atlas, arXiv preprint arXiv:2103.11067, 2021. URL https://arxiv.org/abs/2103.11067v1.
[914] Alex Roxin and Stefano Fusi. Efficient partitioning of memory systems and its importance for memory consolidation. PLoS Comput. Biol., 2013.
[915] Kaushik Roy, Yuxin Zi, Vignesh Narayanan, Manas Gaur, and Amit P. Sheth. Knowledge-infused self attention transformers, arXiv preprint arXiv:2306.13501, 2023. URL https://arxiv.org/abs/ 2306.13501v1.
[916] Dongyu Ru, Lin Qiu, Xiangkun Hu, Tianhang Zhang, Peng Shi, Shuaichen Chang, Jiayang Cheng,
Cunxiang Wang, Shichao Sun, Huanyu Li, Zizhao Zhang, Binjie Wang, Jiarong Jiang, Tong He,
Zhiguo Wang, Pengfei Liu, Yue Zhang, and Zheng Zhang. Ragchecker: A fine-grained framework for
diagnosing retrieval-augmented generation. Neural Information Processing Systems, 2024.
[917] Jingqing Ruan, Yihong Chen, Bin Zhang, Zhiwei Xu, Tianpeng Bao, Guoqing Du, Shiwei Shi, Hangyu
Mao, Ziyue Li, Xingyu Zeng, and Rui Zhao. Tptu: Large language model-based ai agents for task
planning and tool usage, arXiv preprint arXiv:2308.03427, 2023. URL https://arxiv.org/
abs/2308.03427.
[918] M. Russak, Umar Jamil, Christopher Bryant, Kiran Kamble, Axel Magnuson, Mateusz Russak, and
Waseem Alshikh. Writing in the margins: Better inference pattern for long context retrieval, arXiv
preprint arXiv:2408.14906, 2024. URL https://arxiv.org/abs/2408.14906v1.
[919] Hyun Ryu and Eric Kim. Closer look at efficient inference methods: A survey of speculative decoding,
arXiv preprint arXiv:2411.13157, 2024. URL https://arxiv.org/abs/2411.13157v2.
[920] Iman Saberi and Fatemeh Fard. Context-augmented code generation using programming knowl-
edge graphs, arXiv preprint arXiv:2410.18251, 2024. URL https://arxiv.org/abs/2410.
18251v2.
[921] Abdulfattah Safa and Gözde Gül Sahin. A zero-shot open-vocabulary pipeline for dialogue under-
standing. North American Chapter of the Association for Computational Linguistics, 2024.
[922] Alireza Akhavan Safaei, Pegah Saboori, Reza Ramezani, and Mohammadali Nematbakhsh. Kglm-qa:
A novel approach for knowledge graph-enhanced large language models for question answering.
Conference on Information and Knowledge Technology, 2024.
[923] Avirup Saha, Lakshmi Mandal, Balaji Ganesan, Sambit Ghosh, Renuka Sindhgatta, Carlos Eberhardt,
Dan Debrunner, and Sameep Mehta. Sequential api function calling using graphql schema. Conference
on Empirical Methods in Natural Language Processing, 2024.
[924] Pranab Sahoo, Ayush Kumar Singh, Sriparna Saha, Vinija Jain, S. Mondal, and Aman Chadha. A
systematic survey of prompt engineering in large language models: Techniques and applications,
arXiv preprint arXiv:2402.07927, 2024. URL https://arxiv.org/abs/2402.07927v2.
[925] Rana Salama, Jason Cai, Michelle Yuan, Anna Currey, Monica Sunkara, Yi Zhang, and Yas-
sine Benajiba. Meminsight: Autonomous memory augmentation for llm agents, arXiv preprint
arXiv:2503.21760, 2025. URL https://arxiv.org/abs/2503.21760.
[926] Jefferson Salan, Devyn E Smith, Erica S Shafer, and Rachel A Diana. Variation in encoding context
benefits item recognition. Memory & Cognition, 2024.
[927] Alaa Saleh, Sasu Tarkoma, Praveen Kumar Donta, Naser Hossein Motlagh, S. Dustdar, Susanna
Pirttikangas, and Lauri Lov’en. Usercentrix: An agentic memory-augmented ai framework for
smart spaces, arXiv preprint arXiv:2505.00472, 2025. URL https://arxiv.org/abs/2505.
00472v1.
[928] Leonard Salewski, Stephan Alaniz, Isabel Rio-Torto, Eric Schulz, and Zeynep Akata. In-context
impersonation reveals large language models’ strengths and biases. Neural Information Processing
Systems, 2023.
[929] A. Samsonovich. Toward a unified catalog of implemented cognitive architectures. Biologically
Inspired Cognitive Architectures, 2010.
[930] Narendra Reddy Sanikommu. Model context protocol: Enhancing llm performance for observability
and analytics. European journal of computer science and information technology, 2025.
[931] S. Santhanam. Context based text-generation using lstm networks, arXiv preprint arXiv:2005.00048,
2020. URL https://arxiv.org/abs/2005.00048v1.
[932] G. Santos, Rita Maria Silva Julia, and Marcelo Zanchetta do Nascimento. Diverse prompts: Illumi-
nating the prompt space of large language models with map-elites. IEEE Congress on Evolutionary
Computation, 2025.
[933] Ranjan Sapkota, Konstantinos I. Roumeliotis, and Manoj Karkee. Ai agents vs. agentic ai: A
conceptual taxonomy, applications and challenges, arXiv preprint arXiv:2505.10468, 2025. URL
https://arxiv.org/abs/2505.10468v4.
[934] Anjana Sarkar and Soumyendu Sarkar. Survey of llm agent communication with mcp: A software
design pattern centric review, arXiv preprint arXiv:2506.05364, 2025. URL https://arxiv.org/
abs/2506.05364v1.
[935] Soumajyoti Sarkar and Leonard Lausen. Testing the limits of unified sequence to sequence llm
pretraining on diverse table data tasks, arXiv preprint arXiv:2310.00789, 2023. URL https:
//arxiv.org/abs/2310.00789v1.
[936] Parth Sarthi, Salman Abdullah, Aditi Tuli, Shubh Khanna, Anna Goldie, and Christopher D. Manning.
Raptor: Recursive abstractive processing for tree-organized retrieval. International Conference on
Learning Representations, 2024.
[937] Gabriele Sarti. Umberto-mtsa @ accompl-it: Improving complexity and acceptability prediction
with multi-task learning on self-supervised annotations (short paper). International Workshop on
Evaluation of Natural Language and Speech Tools for Italian, 2020.
[938] Apoorv Saxena, Adrian Kochsiek, and Rainer Gemulla. Sequence-to-sequence knowledge graph
completion and question answering. Annual Meeting of the Association for Computational Linguistics,
2022.
[939] Timo Schick, Jane Dwivedi-Yu, Roberto Dessì, R. Raileanu, M. Lomeli, Luke Zettlemoyer, Nicola
Cancedda, and Thomas Scialom. Toolformer: Language models can teach themselves to use tools.
Neural Information Processing Systems, 2023.
[940] Guido Schillaci, Uwe Schmidt, and Luis Miranda. Prediction error-driven memory consolidation for
continual learning: On the case of adaptive greenhouse models. KI - Künstliche Intelligenz, 35(1):
71–80, 2021. ISSN 1610-1987. doi: 10.1007/s13218-020-00700-8. URL http://dx.doi.org/
10.1007/s13218-020-00700-8.
[941] Florian Schneider, Narges Baba Ahmadi, Niloufar Baba Ahmadi, Iris Vogel, Martin Semmann, and
Christian Biemann. Collex - a multimodal agentic rag system enabling interactive exploration of
scientific collections. arXiv preprint, 2025.
[942] Sheila Schoepp, Masoud Jafaripour, Yingyue Cao, Tianpei Yang, Fatemeh Abdollahi, Shadan Golestan,
Zahin Sufiyan, Osmar R. Zaiane, and Matthew E. Taylor. The evolving landscape of llm- and vlm-
integrated reinforcement learning. arXiv preprint, 2025.
[943] Lianlei Shan, Shixian Luo, Zezhou Zhu, Yu Yuan, and Yong Wu. Cognitive memory in large
language models, arXiv preprint arXiv:2504.02441, 2025. URL https://arxiv.org/abs/2504.
02441v2.
[944] Wenbo Shang and Xin Huang. A survey of large language models on generative graph analytics:
Query, learning, and applications, arXiv preprint arXiv:2404.14809v2, 2024. URL https://arxiv.
org/abs/2404.14809v2.
[945] Yunfan Shao, Linyang Li, Junqi Dai, and Xipeng Qiu. Character-llm: A trainable agent for role-playing,
arXiv preprint arXiv:2310.10158, 2023. URL https://arxiv.org/abs/2310.10158.
[946] Yutong Shao and N. Nakashole. On linearizing structured data in encoder-decoder language models:
Insights from text-to-sql. North American Chapter of the Association for Computational Linguistics,
2024.
[947] Zhihong Shao, Yeyun Gong, Yelong Shen, Minlie Huang, Nan Duan, and Weizhu Chen. Synthetic
prompting: Generating chain-of-thought demonstrations for large language models. International
Conference on Machine Learning, 2023.
[948] Peter Shaw, Mandar Joshi, James Cohan, Jonathan Berant, Panupong Pasupat, Hexiang Hu, Urvashi
Khandelwal, Kenton Lee, and Kristina Toutanova. From pixels to ui actions: Learning to follow
instructions via graphical user interfaces. Neural Information Processing Systems, 2023.
[949] Jonathan Shen, Ruoming Pang, Ron J. Weiss, M. Schuster, N. Jaitly, Zongheng Yang, Z. Chen,
Yu Zhang, Yuxuan Wang, R. Skerry-Ryan, R. Saurous, Yannis Agiomyrgiannakis, and Yonghui Wu.
Natural tts synthesis by conditioning wavenet on mel spectrogram predictions. IEEE International
Conference on Acoustics, Speech, and Signal Processing, 2017.
[950] Junhong Shen, Atishay Jain, Zedian Xiao, Ishan Amlekar, Mouad Hadji, Aaron Podolny, and Ameet
Talwalkar. Scribeagent: Towards specialized web agents using production-scale workflow data.
2024.
[951] Junhong Shen, Hao Bai, Lunjun Zhang, Yifei Zhou, Amrith Setlur, Shengbang Tong, Diego Caples,
Nan Jiang, Tong Zhang, Ameet Talwalkar, and Aviral Kumar. Thinking vs. doing: Agents that reason
by scaling test-time interaction, arXiv preprint arXiv:2506.07976, 2025. URL https://arxiv.
org/abs/2506.07976.
[952] Weizhou Shen, Chenliang Li, Fanqi Wan, Shengyi Liao, Shaopeng Lai, Bo Zhang, Yingcheng Shi,
Yuning Wu, Gang Fu, Zhansheng Li, Bin Yang, Ji Zhang, Fei Huang, Jingren Zhou, and Ming Yan.
Qwenlong-cprs: Towards ∞-llms with dynamic context optimization. arXiv preprint, 2025.
[953] Yongliang Shen, Kaitao Song, Xu Tan, Dongsheng Li, Weiming Lu, and Y. Zhuang. Hugginggpt:
Solving ai tasks with chatgpt and its friends in hugging face. Neural Information Processing Systems,
2023.
[954] Zhuocheng Shen. Llm with tools: A survey, arXiv preprint arXiv:2409.18807, 2024. URL https:
//arxiv.org/abs/2409.18807v1.
[955] Ying Sheng, Lianmin Zheng, Binhang Yuan, Zhuohan Li, Max Ryabinin, Daniel Y. Fu, Zhiqiang Xie,
Beidi Chen, Clark W. Barrett, Joseph Gonzalez, Percy Liang, Christopher Ré, Ion Stoica, and Ce Zhang.
High-throughput generative inference of large language models with a single gpu. International
Conference on Machine Learning, 2023.
[956] Dingfeng Shi, Jingyi Cao, Qianben Chen, Weichen Sun, Weizhen Li, Hongxuan Lu, Fangchen Dong,
Tianrui Qin, King Zhu, Minghao Liu, Jian Yang, Ge Zhang, Jiaheng Liu, Changwang Zhang, Jun
Wang, Y. Jiang, and Wangchunshu Zhou. Taskcraft: Automated generation of agentic tasks, arXiv
preprint arXiv:2506.10055, 2025. URL https://arxiv.org/abs/2506.10055v2.
[957] Han Shi, Jiahui Gao, Xiaozhe Ren, Hang Xu, Xiaodan Liang, Zhenguo Li, and J. Kwok. Sparsebert:
Rethinking the importance analysis in self-attention. International Conference on Machine Learning,
2021.
[958] Peng Shi, Patrick Ng, Zhiguo Wang, Henghui Zhu, Alexander Hanbo Li, Jun Wang, C. D. Santos, and
Bing Xiang. Learning contextual representations for semantic parsing with generation-augmented
pre-training. AAAI Conference on Artificial Intelligence, 2020.
[959] Weijia Shi, Xiaochuang Han, M. Lewis, Yulia Tsvetkov, Luke Zettlemoyer, and S. Yih. Trusting your
evidence: Hallucinate less with context-aware decoding. North American Chapter of the Association
for Computational Linguistics, 2023.
[960] Zhengliang Shi, Shen Gao, Xiuyi Chen, Yue Feng, Lingyong Yan, Haibo Shi, Dawei Yin, Zhumin
Chen, Suzan Verberne, and Zhaochun Ren. Tool learning in the wild: Empowering language models
as automatic tool agents. The Web Conference, 2024.
[961] Jay Shim, Grant Kruttschnitt, Alyssa Ma, Daniel Kim, Benjamin Chek, Athul Anand, Kevin Zhu,
and Sean O’Brien. Chain-of-thought augmentation with logit contrast for enhanced reasoning in
language models, arXiv preprint arXiv:2407.03600, 2024. URL https://arxiv.org/abs/2407.
03600v2.
[962] Jiho Shin, Reem Aleithan, Hadi Hemmati, and Song Wang. Retrieval-augmented test generation:
How far are we?, arXiv preprint arXiv:2409.12682, 2024. URL https://arxiv.org/abs/2409.
12682v1.
[963] Seongjin Shin, Sang-Woo Lee, Hwijeen Ahn, Sungdong Kim, Hyoungseok Kim, Boseop Kim,
Kyunghyun Cho, Gichang Lee, W. Park, Jung-Woo Ha, and Nako Sung. On the effect of pre-
training corpora on in-context learning by a large-scale language model. North American Chapter of
the Association for Computational Linguistics, 2022.
[964] Noah Shinn, Federico Cassano, Beck Labash, A. Gopinath, Karthik Narasimhan, and Shunyu Yao.
Reflexion: language agents with verbal reinforcement learning. Neural Information Processing
Systems, 2023.
[965] Fatemeh Shiri, Xiao-Yu Guo, Mona Far, Xin Yu, Reza Haf, and Yuan-Fang Li. An empirical analysis
on spatial reasoning capabilities of large multimodal models. Conference on Empirical Methods in
Natural Language Processing, 2024.
[966] Masoud Shokrnezhad, Hao Yu, T. Taleb, Renwei Li, Kyunghan Lee, Jaeseung Song, and Cedric
Westphal. Toward a dynamic future with adaptable computing and network convergence (acnc).
IEEE Network, 2024.
[967] Connor Shorten, T. Khoshgoftaar, and B. Furht. Text data augmentation for deep learning. Journal
of Big Data, 2021.
[968] Mohit Shridhar, Jesse Thomason, Daniel Gordon, Yonatan Bisk, Winson Han, Roozbeh Mottaghi,
Luke Zettlemoyer, and D. Fox. Alfred: A benchmark for interpreting grounded instructions for
everyday tasks. Computer Vision and Pattern Recognition, 2019.
[969] Altun Shukurlu. Improving deep knowledge tracing via gated architectures and adaptive optimization,
arXiv preprint arXiv:2504.20070, 2025. URL https://arxiv.org/abs/2504.20070v1.
[970] Lynn L Siegel and M. Kahana. A retrieved context account of spacing and repetition effects in free
recall. Journal of Experimental Psychology. Learning, Memory and Cognition, 2014.
[971] Aditi Singh, Abul Ehtesham, Gaurav Kumar Gupta, Nikhil Kumar Chatta, Saket Kumar, and T. T.
Khoei. Exploring prompt engineering: A systematic review with swot analysis, arXiv preprint
arXiv:2410.12843, 2024. URL https://arxiv.org/abs/2410.12843v1.
[972] Anmolika Singh and Yuhang Diao. Leveraging large language models for optimized item categoriza-
tion using unspsc taxonomy. International Journal on Cybernetics & Informatics, 2024.
[973] Joykirat Singh, Raghav Magazine, Yash Pandya, and A. Nambi. Agentic reasoning and tool integration
for llms via reinforcement learning, arXiv preprint arXiv:2505.01441, 2025. URL https://arxiv.
org/abs/2505.01441v1.
[974] Krishnakant Singh, Thanush Navaratnam, Jannik Holmer, Simone Schaub-Meyer, and Stefan Roth.
Is synthetic data all we need? benchmarking the robustness of models trained with synthetic images.
2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2024.
[975] Ramneet Singh, Sathvik Joel, Abhav Mehrotra, Nalin Wadhwa, Ramakrishna Bairi, Aditya Kanade,
and Nagarajan Natarajan. Code researcher: Deep research agent for large systems code and
commit history, arXiv preprint arXiv:2506.11060, 2025. URL https://arxiv.org/abs/2506.
11060v1.
[976] Aarush Sinha and CU Omkumar. Gmlm: Bridging graph neural networks and language models for
heterophilic node classification, arXiv preprint arXiv:2503.05763, 2025. URL https://arxiv.
org/abs/2503.05763v3.
[977] Sanchit Sinha, Yuguang Yue, Victor Soto, Mayank Kulkarni, Jianhua Lu, and Aidong Zhang. Maml-
en-llm: Model agnostic meta-training of llms for improved in-context learning. Knowledge Discovery
and Data Mining, 2024.
[978] Colin Sisate, Alistair Goldfinch, Vincent Waterstone, Sebastian Kingsley, and Mariana Black-
thorn. Contextually entangled gradient mapping for optimized llm comprehension, arXiv preprint
arXiv:2502.00048, 2025. URL https://arxiv.org/abs/2502.00048v1.
[979] Paloma Sodhi, S. R. K. Branavan, Yoav Artzi, and Ryan McDonald. Step: Stacked llm policies for web
actions, arXiv preprint arXiv:2310.03720, 2024. URL https://arxiv.org/abs/2310.03720.
[980] Manthankumar Solanki. Efficient document retrieval with g-retriever. arXiv preprint, 2025.
[981] Karthik Soman, Peter W Rose, John H Morris, Rabia E Akbas, Brett Smith, Braian Peetoom, Catalina
Villouta-Reyes, G. Cerono, Yongmei Shi, Angela Rizk-Jackson, Sharat Israni, Charlotte A. Nelson,
Sui Huang, and Sergio Baranzini. Biomedical knowledge graph-optimized prompt generation for
large language models. Bioinformatics, 2023.
[982] Lilian Some, Wenli Yang, Michael Bain, and Byeong Kang. A comprehensive survey on integrating
large language models with knowledge-based methods. Knowledge-Based Systems, 2025.
[983] Chan Hee Song, Jiaman Wu, Clay Washington, Brian M. Sadler, Wei-Lun Chao, and Yu Su. Llm-
planner: Few-shot grounded planning for embodied agents with large language models. IEEE
International Conference on Computer Vision, 2022.
[984] Huatong Song, Jinhao Jiang, Yingqian Min, Jie Chen, Zhipeng Chen, Wayne Xin Zhao, Lei Fang, and
Ji-Rong Wen. R1-searcher: Incentivizing the search capability in llms via reinforcement learning.
arXiv preprint, 2025.
[985] Woomin Song, Seunghyuk Oh, Sangwoo Mo, Jaehyung Kim, Sukmin Yun, Jung-Woo Ha, and
Jinwoo Shin. Hierarchical context merging: Better long context understanding for pre-trained llms.
International Conference on Learning Representations, 2024.
[986] Woomin Song, Sai Muralidhar Jayanthi, S. Ronanki, Kanthashree Mysore Sathyendra, Jinwoo Shin,
A. Galstyan, Shubham Katiyar, and S. Bodapati. Compress, gather, and recompute: Reforming
long-context processing in transformers, arXiv preprint arXiv:2506.01215, 2025. URL https:
//arxiv.org/abs/2506.01215v1.
[987] Yewei Song, Xunzhu Tang, Cedric Lothritz, Saad Ezzini, Jacques Klein, Tegawend’e F. Bissyand’e,
A. Boytsov, Ulrick Ble, and Anne Goujon. Callnavi, a challenge and empirical study on llm function
calling and routing, arXiv preprint arXiv:2501.05255, 2025. URL https://arxiv.org/abs/
2501.05255v2.
[988] Yueqi Song, Frank Xu, Shuyan Zhou, and Graham Neubig. Beyond browsing: Api-based web agents,
arXiv preprint arXiv:2410.16464, 2025. URL https://arxiv.org/abs/2410.16464.
[989] S. Srinivasa and Jayati Deshmukh. Paradigms of computational agency. Novel Approaches to Informa-
tion Systems Design, 2021.
[990] B. Staresina, R. Henson, N. Kriegeskorte, and Arjen Alink. Episodic reinstatement in the medial
temporal lobe. Journal of Neuroscience, 2012.
[991] T. Staudigl, C. Vollmar, S. Noachtar, and S. Hanslmayr. Temporal-pattern similarity analysis reveals
the beneficial and detrimental effects of context reinstatement on human memory. Journal of
Neuroscience, 2015.
[992] Kaya Stechly, Karthik Valmeekam, and Subbarao Kambhampati. Chain of thoughtlessness? an
analysis of cot in planning. Neural Information Processing Systems, 2024.
[993] R. Sterken and James Ravi Kirkpatrick. Conversational alignment with artificial intelligence in
context. Philosophical Perspectives, 2025.
[994] Paul Stoewer, Achim Schilling, Andreas K. Maier, and Patrick Krauss. Multi-modal cognitive maps
based on neural networks trained on successor representations, arXiv preprint arXiv:2401.01364,
2023. URL https://arxiv.org/abs/2401.01364v1.
[995] Olly Styles, Sam Miller, Patricio Cerda-Mardini, T. Guha, Victor Sanchez, and Bertie Vidgen.
Workbench: a benchmark dataset for agents in a realistic workplace setting, arXiv preprint
arXiv:2405.00823, 2024. URL https://arxiv.org/abs/2405.00823v2.
[996] Guangxin Su, Yifan Zhu, Wenjie Zhang, Hanchen Wang, and Ying Zhang. Bridging large language
models and graph structure learning models for robust representation learning, arXiv preprint
arXiv:2410.12096, 2024. URL https://arxiv.org/abs/2410.12096v1.
[997] Hong Su, Elke A. Rundensteiner, and Murali Mani. Automaton in or out: run-time plan optimization
for xml stream processing. International Symposium on Signal Processing Systems, 2008.
[998] Hongjin Su, Ruoxi Sun, Jinsung Yoon, Pengcheng Yin, Tao Yu, and Sercan Ö. Arık. Learn-by-interact:
A data-centric framework for self-adaptive agents in realistic environments. 2025.
[999] Jinyan Su, Jennifer Healey, Preslav Nakov, and Claire Cardie. Between underthinking and overthink-
ing: An empirical study of reasoning length and correctness in llms, arXiv preprint arXiv:2505.00127,
2025. URL https://arxiv.org/abs/2505.00127v1.
[1000] Weihang Su, Yichen Tang, Qingyao Ai, Zhijing Wu, and Yiqun Liu. Dragin: Dynamic retrieval
augmented generation based on the real-time information needs of large language models. Annual
Meeting of the Association for Computational Linguistics, 2024.
[1001] Xin Su, Man Luo, Kris W Pan, Tien Pei Chou, Vasudev Lal, and Phillip Howard. Sk-vqa: Synthetic
knowledge generation at scale for training context-augmented multimodal llms. arXiv preprint,
2024.
[1002] Budhitama Subagdja and A. Tan. Neural modeling of sequential inferences and learning over episodic
memory. Neurocomputing, 2015.
[1003] Yang Sui, Yu-Neng Chuang, Guanchu Wang, Jiamu Zhang, Tianyi Zhang, Jiayi Yuan, Hongyi
Liu, Andrew Wen, Shaochen Zhong, Hanjie Chen, and Xia Hu. Stop overthinking: A survey on
efficient reasoning for large language models, arXiv preprint arXiv:2503.16419, 2025. URL https:
//arxiv.org/abs/2503.16419v3.
[1004] Chuanneng Sun, Songjun Huang, and D. Pompili. Llm-based multi-agent reinforcement learning:
Current and future directions, arXiv preprint arXiv:2405.11106, 2024. URL https://arxiv.org/
abs/2405.11106v1.
[1005] Hanshi Sun, Zhuoming Chen, Xinyu Yang, Yuandong Tian, and Beidi Chen. Triforce: Lossless
acceleration of long sequence generation with hierarchical speculative decoding, arXiv preprint
arXiv:2404.11912, 2024. URL https://arxiv.org/abs/2404.11912v3.
[1006] Haotian Sun, Yuchen Zhuang, Lingkai Kong, Bo Dai, and Chao Zhang. Adaplanner: Adaptive
planning from feedback with language models. Neural Information Processing Systems, 2023.
[1007] Jiankai Sun, Chuanyang Zheng, E. Xie, Zhengying Liu, Ruihang Chu, Jianing Qiu, Jiaqi Xu, Mingyu
Ding, Hongyang Li, Mengzhe Geng, Yue Wu, Wenhai Wang, Junsong Chen, Zhangyue Yin, Xiaozhe
Ren, Jie Fu, Junxian He, Wu Yuan, Qi Liu, Xihui Liu, Yu Li, Hao Dong, Yu Cheng, Ming Zhang,
P. Heng, Jifeng Dai, Ping Luo, Jingdong Wang, Jingwei Wen, Xipeng Qiu, Yi-Chen Guo, Hui Xiong,
Qun Liu, and Zhenguo Li. A survey of reasoning with foundation models: Concepts, methodologies,
and outlook. ACM Computing Surveys, 2023.
[1008] Jiashuo Sun, Chengjin Xu, Lumingyuan Tang, Sai Wang, Chen Lin, Yeyun Gong, H. Shum, and Jian
Guo. Think-on-graph: Deep and responsible reasoning of large language model with knowledge
graph. arXiv preprint, 2023.
[1009] Lei Sun, Zhengwei Tao, Youdi Li, and Hiroshi Arakawa. Oda: Observation-driven agent for integrating
llms and knowledge graphs, arXiv preprint arXiv:2404.07677, 2024. URL https://arxiv.org/
abs/2404.07677.
[1010] Lei Sun, Xinchen Wang, and Youdi Li. Pyramid-driven alignment: Pyramid principle guided integra-
tion of large language models and knowledge graphs, arXiv preprint arXiv:2410.12298, 2024. URL
https://arxiv.org/abs/2410.12298v2.
[1011] Liangtai Sun, Xingyu Chen, Lu Chen, Tianle Dai, Zichen Zhu, and Kai Yu. Meta-gui: Towards multi-
modal conversational agents on mobile gui. Conference on Empirical Methods in Natural Language
Processing, 2022.
[1012] Lijun Sun, Yijun Yang, Qiqi Duan, Yuhui Shi, Chao Lyu, Yu-Cheng Chang, Chin-Teng Lin, and
Yang Shen. Multi-agent coordination across diverse applications: A survey, arXiv preprint
arXiv:2502.14743, 2025. URL https://arxiv.org/abs/2502.14743v2.
[1013] Qianru Sun, Yaoyao Liu, Tat-Seng Chua, and B. Schiele. Meta-transfer learning for few-shot learning.
Computer Vision and Pattern Recognition, 2018.
[1014] Yu Sun, Shuohuan Wang, Yukun Li, Shikun Feng, Hao Tian, Hua Wu, and Haifeng Wang. Ernie
2.0: A continual pre-training framework for language understanding. AAAI Conference on Artificial
Intelligence, 2019.
[1015] Rao Surapaneni, Miku Jha, Michael Vakoc, and Todd Segal. Announcing the
agent2agent protocol (a2a). https://developers.googleblog.com/en/
a2a-a-new-era-of-agent-interoperability/, April 2025. [Online; accessed 17-
July-2025].
[1016] Stefan Szeider. Mcp-solver: Integrating language models with constraint programming systems.
arXiv preprint, 2024.
[1017] Daniel Szelogowski. Engram memory encoding and retrieval: A neurocomputational perspective,
arXiv preprint arXiv:2506.01659, 2025. URL https://arxiv.org/abs/2506.01659v1.
[1018] N. Taatgen, David Huss, D. Dickison, and John R. Anderson. The acquisition of robust and flexible
cognitive skills. Journal of experimental psychology. General, 2008.
[1019] Jihoon Tack, Jaehyung Kim, Eric Mitchell, Jinwoo Shin, Yee Whye Teh, and Jonathan Richard
Schwarz. Online adaptation of language models with a memory of amortized contexts. Neural
Information Processing Systems, 2024.
[1020] Yan Tai, Weichen Fan, Zhao Zhang, Feng Zhu, Rui Zhao, and Ziwei Liu. Link-context learning for
multimodal llms. Computer Vision and Pattern Recognition, 2023.
[1021] Fahim Tajwar, Yiding Jiang, Abitha Thankaraj, Sumaita Sadia Rahman, J. Z. Kolter, Jeff Schneider,
and Ruslan Salakhutdinov. Training a generally curious agent, arXiv preprint arXiv:2502.17543,
2025. URL https://arxiv.org/abs/2502.17543v3.
[1022] K. Tallam. From autonomous agents to integrated systems, a new paradigm: Orchestrated distributed
intelligence, arXiv preprint arXiv:2503.13754, 2025. URL https://arxiv.org/abs/2503.
13754v2.
[1023] A. Tan, Budhitama Subagdja, Di Wang, and Lei Meng. Self-organizing neural networks for universal
learning and multimodal memory encoding. Neural Networks, 2019.
[1024] Chuanyuan Tan, Yuehe Chen, Wenbiao Shao, and Wenliang Chen. Make a choice! knowledge
base question answering with in-context learning, arXiv preprint arXiv:2305.13972, 2023. URL
https://arxiv.org/abs/2305.13972v1.
[1025] Sijun Tan, Xiuyu Li, Shishir G. Patil, Ziyang Wu, Tianjun Zhang, Kurt Keutzer, Joseph E. Gonzalez,
and Raluca A. Popa. Lloco: Learning long contexts offline. Conference on Empirical Methods in
Natural Language Processing, 2024.
[1026] Xiaoyu Tan, Haoyu Wang, Xihe Qiu, Yuan Cheng, Yinghui Xu, Wei Chu, and Yuan Qi. Struct-x:
Enhancing large language models reasoning with structured data. arXiv preprint, 2024.
[1027] Zhaoxuan Tan and Meng Jiang. User modeling in the era of large language models: Current research
and future directions. IEEE Data Engineering Bulletin, 2023.
[1028] Zhijie Tan, Xu Chu, Weiping Li, and Tong Mo. Order matters: Exploring order sensitivity in
multimodal large language models, arXiv preprint arXiv:2410.16983v1, 2024. URL https://
arxiv.org/abs/2410.16983v1.
[1029] Matthew Tancik, Pratul P. Srinivasan, B. Mildenhall, Sara Fridovich-Keil, N. Raghavan, Utkarsh
Singhal, R. Ramamoorthi, J. Barron, and Ren Ng. Fourier features let networks learn high frequency
functions in low dimensional domains. Neural Information Processing Systems, 2020.
[1030] Fei Tang, Haolei Xu, Hang Zhang, Siqi Chen, Xingyu Wu, Yongliang Shen, Wenqi Zhang, Guiyang
Hou, Zeqi Tan, Yuchen Yan, Kaitao Song, Jian Shao, Weiming Lu, Jun Xiao, and Yueting Zhuang. A
survey on (m)llm-based gui agents, arXiv preprint arXiv:2504.13865, 2025. URL https://arxiv.
org/abs/2504.13865v2.
[1031] Jiabin Tang, Yuhao Yang, Wei Wei, Lei Shi, Lixin Su, Suqi Cheng, Dawei Yin, and Chao Huang.
Graphgpt: Graph instruction tuning for large language models. Annual International ACM SIGIR
Conference on Research and Development in Information Retrieval, 2023.
[1032] Jiabin Tang, Yuhao Yang, Wei Wei, Lei Shi, Lixin Su, Suqi Cheng, Dawei Yin, and Chao Huang.
Graphgpt: Graph instruction tuning for large language models. In Proceedings of the 47th International
ACM SIGIR Conference on Research and Development in Information Retrieval, pages 491–500, 2024.
[1033] Jianheng Tang, Qifan Zhang, Yuhan Li, Nuo Chen, and Jia Li. Grapharena: Evaluating and exploring
large language models on graph computation. arXiv preprint arXiv:2407.00379, 2024.
[1034] Xiangru Tang, Yiming Zong, Jason Phang, Yilun Zhao, Wangchunshu Zhou, Arman Cohan, and Mark
Gerstein. Struc-bench: Are large language models good at generating complex structured tabular
data? North American Chapter of the Association for Computational Linguistics, 2024.
[1035] Xiangru Tang, Tianyu Hu, Muyang Ye, Yanjun Shao, Xunjian Yin, Siru Ouyang, Wangchunshu Zhou,
Pan Lu, Zhuosheng Zhang, Yilun Zhao, Arman Cohan, and Mark Gerstein. Chemagent: Self-updating
library in large language models improves chemical reasoning, arXiv preprint arXiv:2501.06590,
2025. URL https://arxiv.org/abs/2501.06590v1.
[1036] Xuemei Tang, Jun Wang, and Q. Su. Chinese word segmentation with heterogeneous graph neural net-
work, arXiv preprint arXiv:2201.08975, 2022. URL https://arxiv.org/abs/2201.08975v1.
[1037] Yiqing Tang, Xingyuan Dai, Chengchong Zhao, Qi Cheng, and Yisheng Lv. Large language model-
driven urban traffic signal control. Australian and New Zealand Control Conference, 2024.
[1038] Yongjian Tang, Rakebul Hasan, and Thomas Runkler. Fsponer: Few-shot prompt optimization for
named entity recognition in domain-specific scenarios. European Conference on Artificial Intelligence,
2024.
[1039] Yunlong Tang, Daiki Shimada, Jing Bi, Hang Hua, and Chenliang Xu. Empowering llms with pseudo-
untrimmed videos for audio-visual temporal understanding. AAAI Conference on Artificial Intelligence,
2024.
[1040] Yao Tao, Yehui Tang, Yun Wang, Mingjian Zhu, Hailin Hu, and Yunhe Wang. Saliency-driven
dynamic token pruning for large language models, arXiv preprint arXiv:2504.04514, 2025. URL
https://arxiv.org/abs/2504.04514v2.
[1041] Denis Tarasov and Kumar Shridhar. Distilling llms’ decomposition abilities into compact language mod-
els, arXiv preprint arXiv:2402.01812, 2024. URL https://arxiv.org/abs/2402.01812v1.
[1042] Pittawat Taveekitworachai, Potsawee Manakul, Kasima Tharnpipitchai, and Kunat Pipatanakul.
Typhoon t1: An open thai reasoning model, arXiv preprint arXiv:2502.09042, 2025. URL https:
//arxiv.org/abs/2502.09042v2.
[1043] Yi Tay, Anh Tuan Luu, Minh C. Phan, and S. Hui. Multi-task neural network for non-discrete attribute
prediction in knowledge graphs. International Conference on Information and Knowledge Management,
2017.
[1044] 36Kr Editorial Team. The future of ai: From parameter scaling to context scaling. Online, 2025.
URL https://36kr.com/p/3337269379328264. Chinese business and technology media pub-
lication discussing context scaling in large language models.
[1045] Junfeng Tian, Da Zheng, Yang Cheng, Rui Wang, Colin Zhang, and Debing Zhang. Untie the knots:
An efficient data augmentation strategy for long-context pre-training in language models, arXiv
preprint arXiv:2409.04774, 2024. URL https://arxiv.org/abs/2409.04774v1.
[1046] S Tian, R Wang, H Guo, P Wu, Y Dong, and X Wang. . . . Ego-r1: Chain-of-tool-thought for ultra-long
egocentric video reasoning. 2025. URL https://arxiv.org/abs/2506.13654.
[1047] Shulin Tian, Ruiqi Wang, Hongming Guo, Penghao Wu, Yuhao Dong, Xiuying Wang, Jingkang Yang,
Hao Zhang, Hongyuan Zhu, and Ziwei Liu. Ego-r1: Chain-of-tool-thought for ultra-long egocentric
video reasoning. arXiv preprint, 2025.
[1048] Ramine Tinati, Xin Wang, Ian C. Brown, T. Tiropanis, and W. Hall. A streaming real-time web
observatory architecture for monitoring the health of social machines. The Web Conference, 2015.
[1049] Kushal Tirumala, Aram H. Markosyan, Luke Zettlemoyer, and Armen Aghajanyan. Memorization
without overfitting: Analyzing the training dynamics of large language models. Neural Information
Processing Systems, 2022.
[1050] Despina Tomkou, George Fatouros, Andreas Andreou, Georgios Makridis, F. Liarokapis, Dimitrios
Dardanis, Athanasios Kiourtis, John Soldatos, and D. Kyriazis. Bridging industrial expertise and
xr with llm-powered conversational agents, arXiv preprint arXiv:2504.05527, 2025. URL https:
//arxiv.org/abs/2504.05527v1.
[1051] Sabrina Toro, A. V. Anagnostopoulos, Sue Bello, Kai Blumberg, Rhiannon Cameron, Leigh Carmody,
A. Diehl, Damion M. Dooley, William Duncan, P. Fey, Pascale Gaudet, Nomi L. Harris, marcin p.
joachimiak, Leila Kiani, Tiago Lubiana, M. Munoz-Torres, Shawn T. O’Neil, David Osumi-Sutherland,
Aleix Puig, Justin Reese, L. Reiser, Sofia M C Robb, Troy Ruemping, James Seager, Eric Sid, Ray
Stefancsik, Magalie Weber, Valerie Wood, M. Haendel, and Christopher J. Mungall. Dynamic retrieval
augmented generation of ontologies using artificial intelligence (dragon-ai). Journal of Biomedical
Semantics, 2023.
[1052] Fernanda M De La Torre, Cathy Mengying Fang, Han Huang, Andrzej Banburski-Fahey, Judith Amores
Fernandez, and Jaron Lanier. Llmr: Real-time prompting of interactive worlds using large language
models. International Conference on Human Factors in Computing Systems, 2023.
[1053] Martina Toshevska and Sonja Gievska. Llm-based text style transfer: Have we taken a step forward?
IEEE Access, 2025.
[1054] Fouad Trad and Ali Chehab. Evaluating the efficacy of prompt-engineered large multimodal models
versus fine-tuned vision transformers in image-based security applications. ACM Transactions on
Intelligent Systems and Technology, 2024.
[1055] Khanh-Tung Tran, Dung Dao, Minh-Duong Nguyen, Quoc-Viet Pham, Barry O’Sullivan, and Hoang D.
Nguyen. Multi-agent collaboration mechanisms: A survey of llms, arXiv preprint arXiv:2501.06322,
2025. URL https://arxiv.org/abs/2501.06322v1.
[1056] Harold Triedman, Rishi Jha, and Vitaly Shmatikov. Multi-agent systems execute arbitrary malicious
code, arXiv preprint arXiv:2503.12188, 2025. URL https://arxiv.org/abs/2503.12188v1.
[1057] H. Trivedi, Tushar Khot, Mareike Hartmann, R. Manku, Vinty Dong, Edward Li, Shashank Gupta,
Ashish Sabharwal, and Niranjan Balasubramanian. Appworld: A controllable world of apps and people
for benchmarking interactive coding agents. Annual Meeting of the Association for Computational
Linguistics, 2024.
[1058] Yun-Da Tsai, Ting-Yu Yen, Pei-Fu Guo, Zhe-Yan Li, and Shou-De Lin. Text-centric alignment for
multi-modality learning, arXiv preprint arXiv:2402.08086v2, 2024. URL https://arxiv.org/
abs/2402.08086v2.
[1059] Tao Tu, M. Schaekermann, Anil Palepu, Khaled Saab, Jan Freyberg, Ryutaro Tanno, Amy Wang,
Brenna Li, Mohamed Amin, Yong Cheng, Elahe Vedadi, Nenad Tomašev, Shekoofeh Azizi, Karan
Singhal, Le Hou, Albert Webson, Kavita Kulkarni, S. Mahdavi, Christopher Semturs, Juraj Gottweis,
Joelle Barral, Katherine Chou, Greg S. Corrado, Yossi Matias, A. Karthikesalingam, and Vivek
Natarajan. Towards conversational diagnostic artificial intelligence. Nature, 2025.
[1060] Eduard Tulchinskii, Laida Kushnareva, Kristian Kuznetsov, Anastasia Voznyuk, Andrei Andriiainen,
Irina Piontkovskaya, Evgeny Burnaev, and Serguei Barannikov. Listening to the wise few: Select-
and-copy attention heads for multiple-choice qa, arXiv preprint arXiv:2410.02343, 2024. URL
https://arxiv.org/abs/2410.02343v1.
[1061] Meet Udeshi, Minghao Shao, Haoran Xi, Nanda Rani, Kimberly Milner, Venkata Sai Charan Putrevu,
Brendan Dolan-Gavitt, S. K. Shukla, P. Krishnamurthy, F. Khorrami, Ramesh Karri, and Muham-
mad Shafique. D-cipher: Dynamic collaborative intelligent multi-agent system with planner and
heterogeneous executors for offensive security. arXiv preprint, 2025.
[1062] M. Ursino, Nicole Cesaretti, and G. Pirazzini. A model of working memory for encoding multiple
items and ordered sequences exploiting the theta-gamma code. Cognitive Neurodynamics, 2022.
[1063] D. H. V. Uytsel, Filip Van Aelten, and Dirk Van Compernolle. A structured language model based
on context-sensitive probabilistic left-corner parsing. North American Chapter of the Association for
Computational Linguistics, 2001.
[1064] Saeid Ario Vaghefi, Aymane Hachcham, Veronica Grasso, Jiska Manicus, Nakiete Msemo, C. Senni,
and Markus Leippold. Ai for climate finance: Agentic retrieval and multi-step reasoning for early
warning system investments, arXiv preprint arXiv:2504.05104, 2025. URL https://arxiv.org/
abs/2504.05104v2.
[1065] Priyan Vaithilingam, Tianyi Zhang, and Elena L. Glassman. Expectation vs. experience: Evaluating
the usability of code generation tools powered by large language models. CHI Extended Abstracts,
2022.
[1066] Phuc Phan Van, Dat Nguyen Minh, An Dinh Ngoc, and Huy-Phan Thanh. Rx strategist: Prescription
verification using llm agents system, arXiv preprint arXiv:2409.03440, 2024. URL https://arxiv.
org/abs/2409.03440v1.
[1067] Ashish Vaswani, Noam M. Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez,
Lukasz Kaiser, and I. Polosukhin. Attention is all you need. Neural Information Processing Systems,
2017.
[1068] J. D. Velásquez-Henao, Carlos Jaime Franco-Cardona, and Lorena Cadavid-Higuita. Prompt engineer-
ing: a methodology for optimizing interactions with ai-language models in the field of engineering.
DYNA, 2023.
[1069] Gaurav Verma, Rachneet Kaur, Nishan Srishankar, Zhen Zeng, T. Balch, and Manuela Veloso. Adap-
tagent: Adapting multimodal web agents with few-shot learning from human demonstrations, arXiv
preprint arXiv:2411.13451, 2024. URL https://arxiv.org/abs/2411.13451v1.
[1070] Aliaksei Vertsel and Mikhail Rumiantsau. Hybrid llm/rule-based approaches to business insights
generation from structured data, arXiv preprint arXiv:2404.15604, 2024. URL https://arxiv.
org/abs/2404.15604v1.
[1071] Aishwarya Vijayan. A prompt engineering approach for structured data extraction from unstructured
text using conversational llms. International Conference on Advances in Computing and Artificial
Intelligence, 2023.
[1072] Juraj Vladika, Alexander Fichtl, and Florian Matthes. Diversifying knowledge enhancement of
biomedical language models using adapter modules and knowledge graphs. International Conference
on Agents and Artificial Intelligence, 2023.
[1073] James Vo. Sparseaccelerate: Efficient long-context inference for mid-range gpus, arXiv preprint
arXiv:2412.06198, 2024. URL https://arxiv.org/abs/2412.06198v1.
[1074] Blavz vSkrlj, Boshko Koloski, S. Pollak, and Nada Lavravc. From symbolic to neural and back:
Exploring knowledge graph-large language model synergies. arXiv preprint, 2025.
[1075] Tom Völker, Jan Pfister, Tobias Koopmann, and Andreas Hotho. From chat to publication management:
Organizing your related work using bibsonomy & llms. Conference on Human Information Interaction
and Retrieval, 2024.
[1076] D. Walton. Using argumentation schemes to find motives and intentions of a rational agent. Argument
Comput., 2020.
[1077] Hanlong Wan, Jian Zhang, Yan Chen, Weili Xu, and Fan Feng. Generative ai application for building
industry. Building Simulation, 2024.
[1078] Jun Wan and Lingrui Mei. Large language models as computable approximations to solomonoff in-
duction, arXiv preprint arXiv:2505.15784, 2025. URL https://arxiv.org/abs/2505.15784.
[1079] Luanbo Wan and Weizhi Ma. Storybench: A dynamic benchmark for evaluating long-term memory
with multi turns, arXiv preprint arXiv:2506.13356, 2025. URL https://arxiv.org/abs/2506.
13356v1.
[1080] Bernie Wang, Si ting Xu, K. Keutzer, Yang Gao, and Bichen Wu. Improving context-based
meta-reinforcement learning with self-supervised trajectory contrastive learning, arXiv preprint
arXiv:2103.06386, 2021. URL https://arxiv.org/abs/2103.06386v1.
[1081] Bing Wang, Xinnian Liang, Jian Yang, Hui Huang, Shuangzhi Wu, Peihao Wu, Lu Lu, Zejun Ma, and
Zhoujun Li. Scm: Enhancing large language model with self-controlled memory framework, arXiv
preprint arXiv:2304.13343, 2025. URL https://arxiv.org/abs/2304.13343.
[1082] Cangqing Wang, Yutian Yang, Ruisi Li, Dan Sun, Ruicong Cai, Yuzhu Zhang, and Chengqian Fu.
Adapting llms for efficient context processing through soft prompt compression. Proceedings of the
International Conference on Modeling, Natural Language Processing and Machine Learning, 2024.
[1083] Chaozheng Wang, Yuanhang Yang, Cuiyun Gao, Yun Peng, Hongyu Zhang, and Michael R. Lyu.
Prompt tuning in code intelligence: An experimental evaluation. IEEE Transactions on Software
Engineering, 2023.
[1084] Dongsheng Wang, Zhiqiang Ma, Armineh Nourbakhsh, Kang Gu, and Sameena Shah. Docgraphlm:
Documental graph language model for information extraction. Annual International ACM SIGIR
Conference on Research and Development in Information Retrieval, 2023.
[1085] Fan Wang, Chuan Lin, Yang Cao, and Yu Kang. Benchmarking general purpose in-context learning,
arXiv preprint arXiv:2405.17234, 2024. URL https://arxiv.org/abs/2405.17234v6.
[1086] Guanzhi Wang, Yuqi Xie, Yunfan Jiang, Ajay Mandlekar, Chaowei Xiao, Yuke Zhu, Linxi Fan, and
Anima Anandkumar. Voyager: An open-ended embodied agent with large language models, arXiv
preprint arXiv:2305.16291, 2023. URL https://arxiv.org/abs/2305.16291.
[1087] Guoqing Wang, Zeyu Sun, Zhihao Gong, Sixiang Ye, Yizhou Chen, Yifan Zhao, Qing-Lin Liang, and
Dan Hao. Do advanced language models eliminate the need for prompt engineering in software
engineering?, arXiv preprint arXiv:2411.02093, 2024. URL https://arxiv.org/abs/2411.
02093v1.
[1088] Hanlin Wang, Zhan Tong, Kecheng Zheng, Yujun Shen, and Limin Wang. Contextual ad narration
with interleaved multimodal sequence. Computer Vision and Pattern Recognition, 2024.
[1089] Hanrui Wang, Zhekai Zhang, and Song Han. Spatten: Efficient sparse attention architecture
with cascade token and head pruning. International Symposium on High-Performance Computer
Architecture, 2020.
[1090] Haochen Wang, Xiangtai Li, Zilong Huang, Anran Wang, Jiacong Wang, Tao Zhang, Jiani Zheng,
Sule Bai, Zijian Kang, Jiashi Feng, et al. Traceable evidence enhanced visual grounded reasoning:
Evaluation and methodology. arXiv preprint arXiv:2507.07999, 2025.
[1091] Haochun Wang, Chi Liu, Nuwa Xi, Zewen Qiang, Sendong Zhao, Bing Qin, and Ting Liu. Huatuo:
Tuning llama model with chinese medical knowledge, arXiv preprint arXiv:2304.06975, 2023. URL
https://arxiv.org/abs/2304.06975.
[1092] Haoyu Wang, Tong Teng, Tianyu Guo, An Xiao, Duyu Tang, Hanting Chen, and Yunhe Wang.
Unshackling context length: An efficient selective attention approach through query-key compression,
arXiv preprint arXiv:2502.14477, 2025. URL https://arxiv.org/abs/2502.14477v1.
[1093] Heng Wang, Shangbin Feng, Tianxing He, Zhaoxuan Tan, Xiaochuang Han, and Yulia Tsvetkov. Can
language models solve graph problems in natural language? Neural Information Processing Systems,
2023.
[1094] Hengyi Wang, Haizhou Shi, Shiwei Tan, Weiyi Qin, Wenyuan Wang, Tunyu Zhang, A. Nambi,
T. Ganu, and Hao Wang. Multimodal needle in a haystack: Benchmarking long-context capability of
multimodal large language models. North American Chapter of the Association for Computational
Linguistics, 2024.
[1095] Hongru Wang, Cheng Qian, Manling Li, Jiahao Qiu, Boyang Xue, Mengdi Wang, Heng Ji, and Kam-Fai
Wong. Toward a theory of agents as tool-use decision-makers, arXiv preprint arXiv:2506.00886,
2025. URL https://arxiv.org/abs/2506.00886v1.
[1096] Jingjin Wang. Proprag: Guiding retrieval with beam search over proposition paths, arXiv preprint
arXiv:2504.18070, 2025. URL https://arxiv.org/abs/2504.18070v1.
[1097] Jingyu Wang, Lu Zhang, Xueqing Li, Huazhong Yang, and Yongpan Liu. Ulseq-ta: Ultra-long sequence
attention fusion transformer accelerator supporting grouped sparse softmax and dual-path sparse
layernorm. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 2024.
[1098] Jize Wang, Zerun Ma, Yining Li, Songyang Zhang, Cailian Chen, Kai Chen, and Xinyi Le. Gta: A
benchmark for general tool agents. Neural Information Processing Systems, 2024.
[1099] Lei Wang, Chengbang Ma, Xueyang Feng, Zeyu Zhang, Hao ran Yang, Jingsen Zhang, Zhi-Yang
Chen, Jiakai Tang, Xu Chen, Yankai Lin, Wayne Xin Zhao, Zhewei Wei, and Ji rong Wen. A survey
on large language model based autonomous agents. Frontiers Comput. Sci., 2023.
[1100] Lei Wang, Wanyu Xu, Yihuai Lan, Zhiqiang Hu, Yunshi Lan, Roy Ka-Wei Lee, and Ee-Peng Lim. Plan-
and-solve prompting: Improving zero-shot chain-of-thought reasoning by large language models.
Annual Meeting of the Association for Computational Linguistics, 2023.
[1101] Lei Wang, Jingsen Zhang, Hao Yang, Zhiyuan Chen, Jiakai Tang, Zeyu Zhang, Xu Chen, Yankai Lin,
Ruihua Song, Wayne Xin Zhao, et al. When large language model based agent meets user behavior
analysis: A novel user simulation paradigm. 2023.
[1102] Libo Wang. Towards humanoid robot autonomy: A dynamic architecture integrating continuous
thought machines (ctm) and model context protocol (mcp), arXiv preprint arXiv:2505.19339, 2025.
URL https://arxiv.org/abs/2505.19339v1.
[1103] Liya Wang, Jason Chou, Xin Zhou, A. Tien, and Diane M. Baumgartner. Aviationgpt: A large
language model for the aviation domain, arXiv preprint arXiv:2311.17686, 2023. URL https:
//arxiv.org/abs/2311.17686v1.
[1104] Liyuan Wang, Bo Lei, Qian Li, Hang Su, Jun Zhu, and Yi Zhong. Triple-memory networks: A
brain-inspired method for continual learning. IEEE Transactions on Neural Networks and Learning
Systems, 2020.
[1105] Lu Wang, Fangkai Yang, Chaoyun Zhang, Junting Lu, Jiaxu Qian, Shilin He, Pu Zhao, Bo Qiao, Ray
Huang, Si Qin, Qisheng Su, Jiayi Ye, Yudi Zhang, Jian-Guang Lou, Qingwei Lin, Saravan Rajmohan,
Dongmei Zhang, and Qi Zhang. Large action models: From inception to implementation, arXiv
preprint arXiv:2412.10047, 2025. URL https://arxiv.org/abs/2412.10047.
[1106] Peijie Wang, Zhong-Zhi Li, Fei Yin, Xin Yang, Dekang Ran, and Cheng-Lin Liu. Mv-math: Evaluating
multimodal math reasoning in multi-visual contexts. 2025.
[1107] Qineng Wang, Zihao Wang, Ying Su, and Yangqiu Song. On the discussion of large language models:
Symmetry of agents and interplay with prompts, arXiv preprint arXiv:2311.07076, 2023. URL
https://arxiv.org/abs/2311.07076v1.
[1108] Qingyue Wang, Liang Ding, Yanan Cao, Yibing Zhan, Zheng Lin, Shi Wang, Dacheng Tao, and Li Guo.
Divide, conquer, and combine: Mixture of semantic-independent experts for zero-shot dialogue state
tracking, arXiv preprint arXiv:2306.00434, 2023. URL https://arxiv.org/abs/2306.00434.
[1109] Rongzheng Wang, Shuang Liang, Qizhi Chen, Jiasheng Zhang, and Ke Qin. Graphtool-instruction:
Revolutionizing graph reasoning in llms through decomposed subtask instruction. Knowledge Discovery
and Data Mining, 2024.
[1110] Shengnan Wang, Youhui Bai, Lin Zhang, Pingyi Zhou, Shixiong Zhao, Gong Zhang, Sen Wang,
Renhai Chen, Hua Xu, and Hongwei Sun. Xl3m: A training-free framework for llm length extension
based on segment-wise inference, arXiv preprint arXiv:2405.17755, 2024. URL https://arxiv.
org/abs/2405.17755v1.
[1111] Song Wang, Yaochen Zhu, Haochen Liu, Zaiyi Zheng, Chen Chen, and Jundong Li. Knowledge
editing for large language models: A survey. ACM Computing Surveys, 2023.
[1112] Song Wang, Junhong Lin, Xiaojie Guo, Julian Shun, Jundong Li, and Yada Zhu. Reasoning of large
language models over knowledge graphs with super-relations. International Conference on Learning
Representations, 2025.
[1113] Tiannan Wang, Jiamin Chen, Qingrui Jia, Shuai Wang, Ruoyu Fang, Huilin Wang, Zhaowei Gao,
Chunzhao Xie, Chuou Xu, Jihong Dai, Yibin Liu, Jialong Wu, Shengwei Ding, Long Li, Zhiwei Huang,
Xinle Deng, Teng Yu, Gangan Ma, Han Xiao, Z. Chen, Danjun Xiang, Yunxia Wang, Yuanyuan
Zhu, Yichen Xiao, Jing Wang, Yiru Wang, Siran Ding, Jiayang Huang, Jiayi Xu, Yilihamujiang
Tayier, Zhenyu Hu, Yuan Gao, Chengfeng Zheng, Yu-Jie Ye, Yihan Li, Lei Wan, Xinyue Jiang,
Yujie Wang, Siyuan Cheng, Zhule Song, Xiangru Tang, Xiaohua Xu, Ningyu Zhang, Huajun Chen,
Y. Jiang, and Wangchunshu Zhou. Weaver: Foundation models for creative writing, arXiv preprint
arXiv:2401.17268, 2024. URL https://arxiv.org/abs/2401.17268v1.
[1114] Weizhi Wang, Li Dong, Hao Cheng, Haoyu Song, Xiaodong Liu, Xifeng Yan, Jianfeng Gao, and Furu
Wei. Visually-augmented language modeling. International Conference on Learning Representations,
2022.
[1115] Xiang Wang, Tinglin Huang, Dingxian Wang, Yancheng Yuan, Zhenguang Liu, Xiangnan He, and Tat
seng Chua. Learning intents behind interactions with knowledge graph for recommendation. The
Web Conference, 2021.
[1116] Xiao Wang, Isaac Lyngaas, A. Tsaris, Peng Chen, Sajal Dash, Mayanka Chandra Shekar, Tao Luo,
Hong-Jun Yoon, M. Wahib, and J. Gounley. Ultra-long sequence distributed transformer, arXiv
preprint arXiv:2311.02382, 2023. URL https://arxiv.org/abs/2311.02382v2.
[1117] Xiaohan Wang, Yuhui Zhang, Orr Zohar, and S. Yeung-Levy. Videoagent: Long-form video under-
standing with large language model as agent. European Conference on Computer Vision, 2024.
[1118] Xiaolong Wang, Zhaolu Kang, Wangyuxuan Zhai, Xinyue Lou, Yunghwei Lai, Ziyue Wang,
Yawen Wang, Kaiyu Huang, Yile Wang, Peng Li, and Yang Liu. Mucar: Benchmarking multi-
lingual cross-modal ambiguity resolution for multimodal large language models, arXiv preprint
arXiv:2506.17046v1, 2025. URL https://arxiv.org/abs/2506.17046v1.
[1119] Xiaoqiang Wang, Suyuchen Wang, Yun Zhu, and Bang Liu. R3mem: Bridging memory retention and
retrieval via reversible compression. arXiv preprint, 2025.
[1120] Xiaoyang Wang, Pavan Kapanipathi, Ryan Musa, Mo Yu, Kartik Talamadupula, I. Abdelaziz, Maria
Chang, Achille Fokoue, B. Makni, Nicholas Mattei, and M. Witbrock. Improving natural language
inference using external knowledge in the science questions domain. AAAI Conference on Artificial
Intelligence, 2018.
[1121] Xindi Wang, Mahsa Salmani, Parsa Omidi, Xiangyu Ren, Mehdi Rezagholizadeh, and A. Eshaghi.
Beyond the limits: A survey of techniques to extend the context length in large language models.
International Joint Conference on Artificial Intelligence, 2024.
[1122] Xingyao Wang, Yangyi Chen, Lifan Yuan, Yizhe Zhang, Yunzhu Li, Hao Peng, and Heng Ji. Executable
code actions elicit better llm agents. International Conference on Machine Learning, 2024.
[1123] Xuezhi Wang, Jason Wei, D. Schuurmans, Quoc Le, Ed H. Chi, and Denny Zhou. Self-consistency
improves chain of thought reasoning in language models. International Conference on Learning
Representations, 2022.
[1124] Yancheng Wang, Ziyan Jiang, Zheng Chen, Fan Yang, Yingxue Zhou, Eunah Cho, Xing Fan, Xiaojiang
Huang, Yanbin Lu, and Yingzhen Yang. Recmind: Large language model powered agent for rec-
ommendation, arXiv preprint arXiv:2308.14296, 2024. URL https://arxiv.org/abs/2308.
14296.
[1125] Yani Wang. Application of large language models based on knowledge graphs in question-answering
systems: A review. Applied and Computational Engineering, 2024.
[1126] Yanlin Wang, Wanjun Zhong, Yanxian Huang, Ensheng Shi, Min Yang, Jiachi Chen, Hui Li, Yuchi Ma,
Qianxiang Wang, and Zibin Zheng. Agents in software engineering: Survey, landscape, and vision,
arXiv preprint arXiv:2409.09030, 2024. URL https://arxiv.org/abs/2409.09030v2.
[1127] Yaqi Wang and Haipei Xu. Srsa: A cost-efficient strategy-router search agent for real-world human-
machine interactions. 2024 IEEE International Conference on Data Mining Workshops (ICDMW),
2024.
[1128] Yi Wang, Xinhao Li, Ziang Yan, Yinan He, Jiashuo Yu, Xiangyun Zeng, Chenting Wang, Changlian
Ma, Haian Huang, Jianfei Gao, Min Dou, Kaiming Chen, Wenhai Wang, Yu Qiao, Yali Wang, and
Limin Wang. Internvideo2.5: Empowering video mllms with long and rich context modeling. arXiv
preprint, 2025.
[1129] Yiming Wang, Zhuosheng Zhang, and Rui Wang. Element-aware summarization with large language
models: Expert-aligned evaluation and chain-of-thought method. Annual Meeting of the Association
for Computational Linguistics, 2023.
[1130] Yingming Wang and Pepa Atanasova. Self-critique and refinement for faithful natural language
explanations, arXiv preprint arXiv:2505.22823, 2025. URL https://arxiv.org/abs/2505.
22823v1.
[1131] Yiwei Wang, Wei Wang, Yuxuan Liang, Yujun Cai, and Bryan Hooi. Mixup for node and graph
classification. In Proceedings of the Web Conference 2021, pages 3663–3674, 2021.
[1132] Yu Wang, Yifan Gao, Xiusi Chen, Haoming Jiang, Shiyang Li, Jingfeng Yang, Qingyu Yin, Zheng Li,
Xian Li, Bing Yin, Jingbo Shang, and Julian McAuley. Memoryllm: Towards self-updatable large
language models, arXiv preprint arXiv:2402.04624, 2024. URL https://arxiv.org/abs/2402.
04624.
[1133] Yu Wang, Dmitry Krotov, Yuanzhe Hu, Yifan Gao, Wangchunshu Zhou, Julian McAuley, Dan Gut-
freund, Rogério Feris, and Zexue He. M+: Extending memoryllm with scalable long-term memory.
arXiv preprint, 2025.
[1134] Yubin Wang, Xinyang Jiang, De Cheng, Wenli Sun, Dongsheng Li, and Cairong Zhao. Hpt++:
Hierarchically prompting vision-language models with multi-granularity knowledge generation and
improved structure modeling. arXiv preprint, 2024.
[1135] Yujie Wang, Shiju Wang, Shenhan Zhu, Fangcheng Fu, Xinyi Liu, Xuefeng Xiao, Huixia Li, Jiashi Li,
Faming Wu, and Bin Cui. Flexsp: Accelerating large language model training via flexible sequence
parallelism. International Conference on Architectural Support for Programming Languages and
Operating Systems, 2024.
[1136] Yuntao Wang, Yanghe Pan, Zhou Su, Yi Deng, Quan Zhao, L. Du, Tom H. Luan, Jiawen Kang, and
D. Niyato. Large model based agents: State-of-the-art, cooperation paradigms, security and privacy,
and future trends. IEEE Communications Surveys & Tutorials, 2024.
[1137] Yuntao Wang, Shaolong Guo, Yanghe Pan, Zhou Su, Fahao Chen, Tom H. Luan, Peng Li, Jiawen Kang,
and Dusit Niyato. Internet of agents: Fundamentals, applications, and challenges, arXiv preprint
arXiv:2505.07176, 2025. URL https://arxiv.org/abs/2505.07176v1.
[1138] Yuxiang Wang, Xinnan Dai, Wenqi Fan, and Yao Ma. Exploring graph tasks with pure llms: A
comprehensive benchmark and investigation, arXiv preprint arXiv:2502.18771v1, 2025. URL
https://arxiv.org/abs/2502.18771v1.
[1139] Z. Wang, King Zhu, Chunpu Xu, Wangchunshu Zhou, Jiaheng Liu, Yibo Zhang, Jiashuo Wang, Ning
Shi, Siyu Li, Yizhi Li, Haoran Que, Zhaoxiang Zhang, Yuanxing Zhang, Ge Zhang, Ke Xu, Jie Fu, and
Wenhao Huang. Mio: A foundation model on multimodal tokens, arXiv preprint arXiv:2409.17692v3,
2024. URL https://arxiv.org/abs/2409.17692v3.
[1140] Zheng Wang, Shu Xian Teo, Jieer Ouyang, Yongjun Xu, and Wei Shi. M-rag: Reinforcing large
language model performance through retrieval-augmented generation with multiple partitions.
Annual Meeting of the Association for Computational Linguistics, 2024.
[1141] Zhiruo Wang, Zhoujun Cheng, Hao Zhu, Daniel Fried, and Graham Neubig. What are tools anyway?
a survey from the language model perspective, arXiv preprint arXiv:2403.15452, 2024. URL
https://arxiv.org/abs/2403.15452v1.
[1142] Ziyang Wang, Jianzhou You, Haining Wang, Tianwei Yuan, Shichao Lv, Yang Wang, and Limin Sun.
Honeygpt: Breaking the trilemma in terminal honeypots with large language model, arXiv preprint
arXiv:2406.01882, 2024. URL https://arxiv.org/abs/2406.01882v2.
[1143] Ziyue Wang, Chi Chen, Yiqi Zhu, Fuwen Luo, Peng Li, Ming Yan, Ji Zhang, Fei Huang, Maosong Sun,
and Yang Liu. Browse and concentrate: Comprehending multimodal content via prior-llm context
fusion. Annual Meeting of the Association for Computational Linguistics, 2024.
[1144] Zora Zhiruo Wang, Jiayuan Mao, Daniel Fried, and Graham Neubig. Agent workflow memory, arXiv
preprint arXiv:2409.07429, 2024. URL https://arxiv.org/abs/2409.07429.
[1145] Irene Weber. Large language models are pattern matchers: Editing semi-structured and structured
documents with chatgpt. AKWI Jahrestagung, 2024.
[1146] Hui Wei, Chenyue Feng, and Jianning Zhang. Modeling of memory mechanisms in cerebral cortex
and simulation of storage performance, arXiv preprint arXiv:2401.00381, 2023. URL https:
//arxiv.org/abs/2401.00381v2.
[1147] Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Ed H. Chi, F. Xia, Quoc Le, and Denny
Zhou. Chain of thought prompting elicits reasoning in large language models. Neural Information
Processing Systems, 2022.
[1148] Jerry W. Wei, Le Hou, Andrew Kyle Lampinen, Xiangning Chen, Da Huang, Yi Tay, Xinyun Chen,
Yifeng Lu, Denny Zhou, Tengyu Ma, and Quoc V. Le. Symbol tuning improves in-context learning in
language models. Conference on Empirical Methods in Natural Language Processing, 2023.
[1149] Shaopeng Wei, Yu Zhao, Xingyan Chen, Qing Li, Fuzhen Zhuang, Ji Liu, and Gang Kou. Graph
learning and its advancements on large language models: A holistic survey, arXiv preprint
arXiv:2212.08966, 2022. URL https://arxiv.org/abs/2212.08966v5.
[1150] Zhepei Wei, Wenlin Yao, Yao Liu, Weizhi Zhang, Qin Lu, Liang Qiu, Changlong Yu, Puyang Xu, Chao
Zhang, Bing Yin, Hyokun Yun, and Lihong Li. Webagent-r1: Training web agents via end-to-end
multi-turn reinforcement learning. arXiv preprint, 2025.
145
[1151] Zhiyuan Wei, Jing Sun, Zijian Zhang, and Xianhao Zhang. Llm-smartaudit: Advanced smart contract
vulnerability detection. arXiv preprint, 2024.
[1152] Rebecca Westhäußer, Frederik Berenz, Wolfgang Minker, and Sebastian Zepf. Caim: Development
and evaluation of a cognitive ai memory framework for long-term interaction with intelligent agents.
arXiv preprint, 2025.
[1153] Danny Weyns and F. Oquendo. An architectural style for self-adaptive multi-agent systems, arXiv
preprint arXiv:1909.03475, 2019. URL https://arxiv.org/abs/1909.03475v1.
[1154] Erik Wijmans, Brody Huval, Alexander Hertzberg, V. Koltun, and Philipp Krähenbühl. Cut your losses
in large-vocabulary language models. International Conference on Learning Representations, 2024.
[1155] Wikipedia contributors. Agent communications language — Wikipedia, the free encyclopedia,
2025. URL https://en.wikipedia.org/wiki/Agent_Communications_Language. [On-
line; accessed 17-July-2025].
[1156] Genta Indra Winata, Lingjue Xie, Karthik Radhakrishnan, Shijie Wu, Xisen Jin, Pengxiang Cheng,
Mayank Kulkarni, and Daniel Preotiuc-Pietro. Overcoming catastrophic forgetting in massively
multilingual continual learning, arXiv preprint arXiv:2305.16252, 2023. URL https://arxiv.
org/abs/2305.16252.
[1157] Beong woo Kwak, Minju Kim, Dongha Lim, Hyungjoo Chae, Dongjin Kang, Sunghwan Kim, Dongil
Yang, and Jinyoung Yeo. Toolhaystack: Stress-testing tool-augmented language models in realistic
long-term interactions, arXiv preprint arXiv:2505.23662, 2025. URL https://arxiv.org/abs/
2505.23662v1.
[1158] Biao Wu, Yanda Li, Meng Fang, Zirui Song, Zhiwei Zhang, Yunchao Wei, and Ling Chen. Foundations
and recent trends in multimodal mobile agents: A survey, arXiv preprint arXiv:2411.02006, 2024.
URL https://arxiv.org/abs/2411.02006v2.
[1159] Cheng-Kuang Wu, Zhi Rui Tam, Chieh-Yen Lin, Yun-Nung Chen, and Hung yi Lee. Streambench:
Towards benchmarking continuous improvement of language agents. Neural Information Processing
Systems, 2024.
[1160] Jialong Wu, Baixuan Li, Runnan Fang, Wenbiao Yin, Liwen Zhang, Zhengwei Tao, Dingchu Zhang,
Zekun Xi, Yong Jiang, Pengjun Xie, Fei Huang, and Jingren Zhou. Webdancer: Towards autonomous
information seeking agency, arXiv preprint arXiv:2505.22648, 2025. URL https://arxiv.org/
abs/2505.22648v2.
[1161] Jialong Wu, Wenbiao Yin, Yong Jiang, Zhenglin Wang, Zekun Xi, Runnan Fang, Deyu Zhou,
Pengjun Xie, and Fei Huang. Webwalker: Benchmarking llms in web traversal, arXiv preprint
arXiv:2501.07572, 2025. URL https://arxiv.org/abs/2501.07572v2.
[1162] Junde Wu, Jiayuan Zhu, and Yuyuan Liu. Agentic reasoning: Reasoning llms with tools for the
deep research, arXiv preprint arXiv:2502.04644, 2025. URL https://arxiv.org/abs/2502.
04644v1.
[1163] Likang Wu, Zhilan Zheng, Zhaopeng Qiu, Hao Wang, Hongchao Gu, Tingjia Shen, Chuan Qin, Chen
Zhu, Hengshu Zhu, Qi Liu, Hui Xiong, and Enhong Chen. A survey on large language models for
recommendation. World wide web (Bussum), 2023.
[1164] M Wu, J Yang, J Jiang, M Li, K Yan, and H Yu. . . . Vtool-r1: Vlms learn to think with images via
reinforcement learning on multimodal tool use. 2025. URL https://arxiv.org/abs/2505.
19255.
[1165] Mengsong Wu, Tong Zhu, Han Han, Chuanyuan Tan, Xiang Zhang, and Wenliang Chen. Seal-tools:
Self-instruct tool learning dataset for agent tuning and detailed benchmark. Natural Language
Processing and Chinese Computing, 2024.
[1166] Panlong Wu, Ting Wang, Yifei Zhong, Haoqi Zhang, Zitong Wang, and Fangxin Wang. Deep-
form: Reasoning large language model for communication system formulation, arXiv preprint
arXiv:2506.08551, 2025. URL https://arxiv.org/abs/2506.08551v2.
[1167] Qingyun Wu, Gagan Bansal, Jieyu Zhang, Yiran Wu, Beibin Li, Erkang Zhu, Li Jiang, Xiaoyun Zhang,
Shaokun Zhang, Jiale Liu, A. Awadallah, Ryen W. White, Doug Burger, and Chi Wang. Autogen:
Enabling next-gen llm applications via multi-agent conversation, arXiv preprint arXiv:2308.08155,
2023. URL https://arxiv.org/abs/2308.08155v2.
[1168] Ruofan Wu, Youngwon Lee, Fan Shu, Danmei Xu, Seung won Hwang, Zhewei Yao, Yuxiong He,
and Feng Yan. Composerag: A modular and composable rag for corpus-grounded multi-hop ques-
tion answering, arXiv preprint arXiv:2506.00232, 2025. URL https://arxiv.org/abs/2506.
00232v1.
[1169] Shangyu Wu, Ying Xiong, Yufei Cui, Haolun Wu, Can Chen, Ye Yuan, Lianming Huang, Xue Liu,
Tei-Wei Kuo, Nan Guan, and C. Xue. Retrieval-augmented generation for natural language process-
ing: A survey, arXiv preprint arXiv:2407.13193, 2024. URL https://arxiv.org/abs/2407.
13193v3.
[1170] Shirley Wu, Shiyu Zhao, Qian Huang, Kexin Huang, Michihiro Yasunaga, V. Ioannidis, Karthik
Subbian, J. Leskovec, and James Zou. Avatar: Optimizing llm agents for tool usage via contrastive
reasoning. Neural Information Processing Systems, 2024.
[1171] Suhang Wu, Minlong Peng, Yue Chen, Jinsong Su, and Mingming Sun. Eva-kellm: A new benchmark
for evaluating knowledge editing of llms, arXiv preprint arXiv:2308.09954, 2023. URL https:
//arxiv.org/abs/2308.09954.
[1172] Tianhao Wu, Weizhe Yuan, Olga Golovneva, Jing Xu, Yuandong Tian, Jiantao Jiao, Jason E Weston,
and Sainbayar Sukhbaatar. Meta-rewarding language models: Self-improving alignment with
llm-as-a-meta-judge. arXiv preprint, 2024.
[1173] Tong Wu, Chong Xiang, Jiachen T. Wang, and Prateek Mittal. Effectively controlling reasoning
models through thinking intervention, arXiv preprint arXiv:2503.24370, 2025. URL https://
arxiv.org/abs/2503.24370v3.
[1174] Xinbo Wu and L. Varshney. A meta-learning perspective on transformers for causal language modeling.
Annual Meeting of the Association for Computational Linguistics, 2023.
[1175] Xue Wu and Kostas Tsioutsiouliklis. Thinking with knowledge graphs: Enhancing llm reasoning
through structured data, arXiv preprint arXiv:2412.10654, 2024. URL https://arxiv.org/
abs/2412.10654v1.
[1176] Yaxiong Wu, Sheng Liang, Chen Zhang, Yichao Wang, Yongyue Zhang, Huifeng Guo, Ruiming Tang,
and Yong Liu. From human memory to ai memory: A survey on memory mechanisms in the era of
llms, arXiv preprint arXiv:2504.15965, 2025. URL https://arxiv.org/abs/2504.15965v2.
[1177] Zengqing Wu and Takayuki Ito. The hidden strength of disagreement: Unraveling the consensus-
diversity tradeoff in adaptive multi-agent systems, arXiv preprint arXiv:2502.16565, 2025. URL
https://arxiv.org/abs/2502.16565v2.
[1178] Zihao Wu, Lu Zhang, Chao-Yang Cao, Xiao-Xing Yu, Haixing Dai, Chong-Yi Ma, Zheng Liu, Lin Zhao,
Gang Li, Wei Liu, Quanzheng Li, Dinggang Shen, Xiang Li, Dajiang Zhu, and Tianming Liu. Exploring
the trade-offs: Unified large language models vs local fine-tuned models for highly-specific radiology
nli task. IEEE Transactions on Big Data, 2023.
[1179] Zhiheng Xi, Wenxiang Chen, Xin Guo, Wei He, Yiwen Ding, Boyang Hong, Ming Zhang, Junzhe
Wang, Senjie Jin, Enyu Zhou, Rui Zheng, Xiaoran Fan, Xiao Wang, Limao Xiong, Qin Liu, Yuhao
Zhou, Weiran Wang, Changhao Jiang, Yicheng Zou, Xiangyang Liu, Zhangyue Yin, Shihan Dou,
Rongxiang Weng, Wensen Cheng, Qi Zhang, Wenjuan Qin, Yongyan Zheng, Xipeng Qiu, Xuanjing
Huan, and Tao Gui. The rise and potential of large language model based agents: A survey, arXiv
preprint arXiv:2309.07864, 2023. URL https://arxiv.org/abs/2309.07864v3.
[1180] Menglin Xia, Victor Ruehle, Saravan Rajmohan, and Reza Shokri. Minerva: A programmable
memory test benchmark for language models, arXiv preprint arXiv:2502.03358, 2025. URL https:
//arxiv.org/abs/2502.03358v2.
[1181] Yuchen Xia, Manthan Shenoy, N. Jazdi, and M. Weyrich. Towards autonomous system: flexible mod-
ular production system enhanced with large language model agents. IEEE International Conference
on Emerging Technologies and Factory Automation, 2023.
[1182] Yutong Xia, Ao Qu, Yunhan Zheng, Yihong Tang, Dingyi Zhuang, Yuxuan Liang, Cathy Wu, Roger
Zimmermann, and Jinhua Zhao. Reimagining urban science: Scaling causal inference with large
language models, arXiv preprint arXiv:2504.12345v3, 2025. URL https://arxiv.org/abs/
2504.12345v3.
[1183] Zhishang Xiang, Chuanjie Wu, Qinggang Zhang, Shengyuan Chen, Zijin Hong, Xiao Huang, and
Jinsong Su. When to use graphs in rag: A comprehensive analysis for graph retrieval-augmented gener-
ation, arXiv preprint arXiv:2506.05690, 2025. URL https://arxiv.org/abs/2506.05690v1.
[1184] Chaojun Xiao, Pengle Zhang, Xu Han, Guangxuan Xiao, Yankai Lin, Zhengyan Zhang, Zhiyuan Liu,
Song Han, and Maosong Sun. Infllm: Training-free long-context extrapolation for llms with an
efficient context memory. Neural Information Processing Systems, 2024.
[1185] Guangxuan Xiao, Yuandong Tian, Beidi Chen, Song Han, and Mike Lewis. Efficient streaming
language models with attention sinks. International Conference on Learning Representations, 2023.
[1186] MiniCPM Team Chaojun Xiao, Yuxuan Li, Xu Han, Yuzhuo Bai, Jie Cai, Haotian Chen, Wentong Chen,
Xin Cong, Ganqu Cui, Ning Ding, Shengda Fan, Yewei Fang, Zixuan Fu, Wenyu Guan, Yitong Guan,
Junshao Guo, Yu-Xuan Han, Bingxiang He, Yuxian Huang, Cunliang Kong, Qiu-Tong Li, Siyuan Li,
Wenhao Li, Yanghao Li, Yishan Li, Zhen Li, Dan Liu, Biyuan Lin, Yankai Lin, Xiang Long, Quanyu Lu,
Ya-Ting Lu, Pei Luo, Hongya Lyu, Litu Ou, Yinxu Pan, Zekai Qu, Qundong Shi, Zijun Song, Jiayu Su,
Zhou Su, Ao Sun, Xiang ping Sun, Peijun Tang, Fang-Ming Wang, Feng Wang, Shuo Wang, Yudong
Wang, Yesai Wu, Zhenyu Xiao, Jie Xie, Zi-Kang Xie, Yukun Yan, Jia-Li Yuan, Kai Zhang, Lei Zhang,
Linyu Zhang, Xueren Zhang, Yudi Zhang, Hengyu Zhao, Weilin Zhao, Weilun Zhao, Yuanqian Zhao,
Zhijun Zheng, Ge Zhou, Jie Zhou, Wei Zhou, Zihan Zhou, Zi-An Zhou, Zhiyuan Liu, Guoyang Zeng,
Chaochao Jia, Dahai Li, and Maosong Sun. Minicpm4: Ultra-efficient llms on end devices, arXiv
preprint arXiv:2506.07900, 2025. URL https://arxiv.org/abs/2506.07900v1.
[1187] Yang Xiao, Jiashuo Wang, Ruifeng Yuan, Chunpu Xu, Kaishuai Xu, Wenjie Li, and Pengfei Liu. Limopro:
Reasoning refinement for efficient and effective test-time scaling, arXiv preprint arXiv:2505.19187,
2025. URL https://arxiv.org/abs/2505.19187v1.
[1188] Yilin Xiao, Chuang Zhou, Qinggang Zhang, Bo Li, Qing Li, and Xiao Huang. Reliable reason-
ing path: Distilling effective guidance for llm reasoning with knowledge graphs, arXiv preprint
arXiv:2506.10508, 2025. URL https://arxiv.org/abs/2506.10508v1.
[1189] Jian Xie, Kai Zhang, Jiangjie Chen, Tinghui Zhu, Renze Lou, Yuandong Tian, Yanghua Xiao, and
Yu Su. Travelplanner: A benchmark for real-world planning with language agents. International
Conference on Machine Learning, 2024.
[1190] Yuxi Xie, Anirudh Goyal, Xiaobao Wu, Xunjian Yin, Xiao Xu, Min-Yen Kan, Liangming Pan, and
William Yang Wang. Coral: Order-agnostic language modeling for efficient iterative refinement,
arXiv preprint arXiv:2410.09675, 2024. URL https://arxiv.org/abs/2410.09675v1.
[1191] Yue Xing, Tao Yang, Yijiashun Qi, Minggu Wei, Yu Cheng, and Honghui Xin. Structured memory mech-
anisms for stable context representation in large language models, arXiv preprint arXiv:2505.22921,
2025. URL https://arxiv.org/abs/2505.22921v1.
[1192] Guangzhi Xiong, Qiao Jin, Xiao Wang, Yin Fang, Haolin Liu, Yifan Yang, Fangyuan Chen, Zhixing Song,
Dengyu Wang, Minjia Zhang, Zhiyong Lu, and Aidong Zhang. Rag-gym: Systematic optimization of
language agents for retrieval-augmented generation. arXiv preprint, 2025.
[1193] Haoyi Xiong, Zhiyuan Wang, Xuhong Li, Jiang Bian, Zeke Xie, Shahid Mumtaz, and Laura E.
Barnes. Converging paradigms: The synergy of symbolic and connectionist ai in llm-empowered
autonomous agents, arXiv preprint arXiv:2407.08516, 2024. URL https://arxiv.org/abs/
2407.08516v5.
[1194] Junjie Xiong, Changjia Zhu, Shuhang Lin, Chong Zhang, Yongfeng Zhang, Yao Liu, and Lingyao Li. In-
visible prompts, visible threats: Malicious font injection in external resources for large language mod-
els, arXiv preprint arXiv:2505.16957, 2025. URL https://arxiv.org/abs/2505.16957v1.
[1195] Zhen Xiong, Yujun Cai, Bryan Hooi, Nanyun Peng, Zhecheng Li, and Yiwei Wang. Enhancing llm
character-level manipulation via divide and conquer. 2025.
[1196] Zhen Xiong, Yujun Cai, Zhecheng Li, and Yiwei Wang. Mapping the minds of llms: A graph-based
analysis of reasoning llm. 2025.
[1197] Zhen Xiong, Yujun Cai, Zhecheng Li, and Yiwei Wang. Unveiling the potential of diffusion large
language model in controllable generation. 2025.
[1198] Zidi Xiong, Yuping Lin, Wenya Xie, Pengfei He, Jiliang Tang, Himabindu Lakkaraju, and Zhen Xiang.
How memory management impacts llm agents: An empirical study of experience-following behavior,
arXiv preprint arXiv:2505.16067, 2025. URL https://arxiv.org/abs/2505.16067v1.
[1199] Chunmei Xu, Shengheng Liu, Cheng Zhang, Yongming Huang, Zhaohua Lu, and Luxi Yang. Multi-
agent reinforcement learning based distributed transmission in collaborative cloud-edge systems.
IEEE Transactions on Vehicular Technology, 2021.
[1200] Haotian Xu, Xing Wu, Weinong Wang, Zhongzhi Li, Da Zheng, Boyuan Chen, Yi Hu, Shijia Kang,
Jiaming Ji, Yingying Zhang, et al. Redstar: Does scaling long-cot data unlock better slow-reasoning
systems? 2025.
[1201] Hongshen Xu, Su Zhu, Zihan Wang, Hang Zheng, Da Ma, Ruisheng Cao, Shuai Fan, Lu Chen, and
Kai Yu. Reducing tool hallucination via reliability alignment, arXiv preprint arXiv:2412.04141, 2024.
URL https://arxiv.org/abs/2412.04141v3.
[1202] Hu Xu, Gargi Ghosh, Po-Yao (Bernie) Huang, Dmytro Okhonko, Armen Aghajanyan, and Florian
Metze Luke Zettlemoyer Christoph Feichtenhofer. Videoclip: Contrastive pre-training for zero-shot
video-text understanding. Conference on Empirical Methods in Natural Language Processing, 2021.
[1203] Mengjia Xu. Understanding graph embedding methods and their applications. SIAM Review, 2020.
[1204] Minrui Xu, Hongyang Du, Dusist Niyato, Jiawen Kang, Zehui Xiong, Shiwen Mao, Zhu Han, A. Ja-
malipour, Dong In Kim, X. Shen, Victor C. M. Leung, and H. Poor. Unleashing the power of edge-cloud
generative ai in mobile networks: A survey of aigc services. IEEE Communications Surveys and Tuto-
rials, 2023.
[1205] Minrui Xu, D. Niyato, and Christopher G. Brinton. Serving long-context llms at the mobile edge:
Test-time reinforcement learning-based model caching and inference offloading, arXiv preprint
arXiv:2501.14205, 2025. URL https://arxiv.org/abs/2501.14205v1.
[1206] Nan Xu, Fei Wang, Sheng Zhang, Hoifung Poon, and Muhao Chen. From introspection to best
practices: Principled analysis of demonstrations in multimodal in-context learning. North American
Chapter of the Association for Computational Linguistics, 2024.
[1207] Shuhang Xu and Fangwei Zhong. Comet: Metaphor-driven covert communication for multi-agent
language games, arXiv preprint arXiv:2505.18218, 2025. URL https://arxiv.org/abs/2505.
18218v1.
[1208] Tianyang Xu, Haojie Zheng, Chengze Li, Haoxiang Chen, Yixin Liu, Ruoxi Chen, and Lichao Sun.
Noderag: Structuring graph-based rag with heterogeneous nodes, arXiv preprint arXiv:2504.11544,
2025. URL https://arxiv.org/abs/2504.11544v1.
[1209] Wenda Xu, Guanglei Zhu, Xuandong Zhao, Liangming Pan, Lei Li, and W. Wang. Pride and prejudice:
Llm amplifies self-bias in self-refinement. Annual Meeting of the Association for Computational
Linguistics, 2024.
[1210] Wenrui Xu and Keshab K. Parhi. A survey of attacks on large language models, arXiv preprint
arXiv:2505.12567, 2025. URL https://arxiv.org/abs/2505.12567v1.
[1211] Wujiang Xu, Zujie Liang, Kai Mei, Hang Gao, Juntao Tan, and Yongfeng Zhang. A-mem: Agentic
memory for llm agents. arXiv preprint, 2025.
[1212] Wujiang Xu, Kai Mei, Hang Gao, Juntao Tan, Zujie Liang, and Yongfeng Zhang. A-mem: Agentic
memory for llm agents, arXiv preprint arXiv:2502.12110, 2025. URL https://arxiv.org/abs/
2502.12110.
[1213] Yifei Xu, Jingqiao Zhang, Ru He, Liangzhu Ge, Chao Yang, Cheng Yang, and Ying Wu. Sas: Self-
augmentation strategy for language model pre-training. AAAI Conference on Artificial Intelligence,
2021.
[1214] Zhe Xu, Daoyuan Chen, Zhenqing Ling, Yaliang Li, and Ying Shen. Mindgym: What matters in
question synthesis for thinking-centric fine-tuning?, arXiv preprint arXiv:2503.09499, 2025. URL
https://arxiv.org/abs/2503.09499v2.
[1215] Zhentao Xu, Mark Jerome Cruz, Matthew Guevara, Tie Wang, Manasi Deshpande, Xiaofeng Wang,
and Zheng Li. Retrieval-augmented generation with knowledge graphs for customer service question
answering. Annual International ACM SIGIR Conference on Research and Development in Information
Retrieval, 2024.
[1216] Eric Xue, Ke Chen, Zeyi Huang, Yuyang Ji, Yong Jae Lee, and Haohan Wang. Improve: Iterative
model pipeline refinement and optimization leveraging llm experts, arXiv preprint arXiv:2502.18530,
2025. URL https://arxiv.org/abs/2502.18530v2.
[1217] Huiyin Xue and Nikolaos Aletras. Pit one against many: Leveraging attention-head embeddings
for parameter-efficient multi-head attention. Conference on Empirical Methods in Natural Language
Processing, 2023.
[1218] Xiangyuan Xue, Zeyu Lu, Di Huang, Zidong Wang, Wanli Ouyang, and Lei Bai. Comfybench:
Benchmarking llm-based agents in comfyui for autonomously designing collaborative ai systems,
arXiv preprint arXiv:2409.01392, 2024. URL https://arxiv.org/abs/2409.01392v2.
[1219] Bingyu Yan, Xiaoming Zhang, Litian Zhang, Lian Zhang, Ziyi Zhou, Dezhuang Miao, and Chaozhuo
Li. Beyond self-talk: A communication-centric survey of llm-based multi-agent systems. arXiv
preprint, 2025.
[1220] Tianqiang Yan and Tiansheng Xu. Refining the responses of llms by themselves, arXiv preprint
arXiv:2305.04039, 2023. URL https://arxiv.org/abs/2305.04039v1.
[1221] Xu Yan, Junliang Du, Lun Wang, Yingbin Liang, Jiacheng Hu, and Bingxing Wang. The synergistic
role of deep learning and neural architecture search in advancing artificial intelligence. 2024
International Conference on Electronics and Devices, Computational Science (ICEDCS), 2024.
[1222] Yibo Yan, Haomin Wen, Siru Zhong, Wei Chen, Haodong Chen, Qingsong Wen, Roger Zimmermann,
and Yuxuan Liang. Urbanclip: Learning text-enhanced urban region profiling with contrastive
language-image pretraining from the web. The Web Conference, 2023.
[1223] Yuchen Yan, Yongliang Shen, Yang Liu, Jin Jiang, Mengdi Zhang, Jian Shao, and Yueting Zhuang.
Inftythink: Breaking the length limits of long-context reasoning in large language models, arXiv
preprint arXiv:2503.06692, 2025. URL https://arxiv.org/abs/2503.06692v3.
[1224] Chengrun Yang, Xuezhi Wang, Yifeng Lu, Hanxiao Liu, Quoc V. Le, Denny Zhou, and Xinyun Chen.
Large language models as optimizers. International Conference on Learning Representations, 2023.
[1225] Hongkang Yang, Zehao Lin, Wenjin Wang, Hao Wu, Zhiyu Li, Bo Tang, Wenqiang Wei, Jinbo Wang,
Zeyun Tang, Shichao Song, Chenyang Xi, Yu Yu, Kai Chen, Feiyu Xiong, Linpeng Tang, and E. Weinan.
Memory3: Language modeling with explicit memory. Journal of Machine Learning, 2024.
[1226] Jianxin Yang. Longqlora: Efficient and effective method to extend context length of large lan-
guage models, arXiv preprint arXiv:2311.04879, 2023. URL https://arxiv.org/abs/2311.
04879v2.
[1227] Jinghan Yang, Shuming Ma, and Furu Wei. Auto-icl: In-context learning without human supervision.
arXiv preprint, 2023.
[1228] Jingkang Yang, Yuhao Dong, Shuai Liu, Bo Li, Ziyue Wang, Chencheng Jiang, Haoran Tan, Ji-
amu Kang, Yuanhan Zhang, Kaiyang Zhou, and Ziwei Liu. Octopus: Embodied vision-language
programmer from environmental feedback. European Conference on Computer Vision, 2023.
[1229] John Yang, Carlos E. Jimenez, Alexander Wettig, Kilian Adriano Lieret, Shunyu Yao, Karthik
Narasimhan, and Ofir Press. Swe-agent: Agent-computer interfaces enable automated software
engineering. Neural Information Processing Systems, 2024.
[1230] Junhan Yang, Zheng Liu, Shitao Xiao, Chaozhuo Li, Defu Lian, Sanjay Agrawal, Amit Singh,
Guangzhong Sun, and Xing Xie. Graphformers: Gnn-nested transformers for representation learning
on textual graph. Neural Information Processing Systems, 2021.
[1231] Ke Yang, Yao Liu, Sapana Chaudhary, Rasool Fakoor, Pratik Chaudhari, George Karypis, and Huzefa
Rangwala. Agentoccam: A simple yet strong baseline for llm-based web agents. 2024.
[1232] Lin F. Yang, Hongyang Chen, Zhao Li, Xiao Ding, and Xindong Wu. Give us the facts: Enhancing
large language models with knowledge graphs for fact-aware language modeling. IEEE Transactions
on Knowledge and Data Engineering, 2023.
[1233] Ling Yang, Ye Tian, Bowen Li, Xinchen Zhang, Ke Shen, Yunhai Tong, and Mengdi Wang. Mmada:
Multimodal large diffusion language models, arXiv preprint arXiv:2505.15809v1, 2025. URL
https://arxiv.org/abs/2505.15809v1.
[1234] R Yang, L Song, Y Li, S Zhao, and Y Ge. . . . Gpt4tools: Teaching large language model to use
tools via self-instruction. 2023. URL https://proceedings.neurips.cc/paper_files/
paper/2023/hash/e393677793767624f2821cec8bdd02f1-Abstract-Conference.
html?utm_campaign=Artificial%2BIntelligence%2BWeekly&utm_medium=email&
utm_source=Artificial_Intelligence_Weekly_411.
[1235] Rui Yang, Lin Song, Yanwei Li, Sijie Zhao, Yixiao Ge, Xiu Li, and Ying Shan. Gpt4tools: Teaching
large language model to use tools via self-instruction. Neural Information Processing Systems, 2023.
[1236] Shang Yang, Junxian Guo, Haotian Tang, Qinghao Hu, Guangxuan Xiao, Jiaming Tang, Yujun
Lin, Zhijian Liu, Yao Lu, and Song Han. Lserve: Efficient long-sequence llm serving with unified
sparse attention, arXiv preprint arXiv:2502.14866, 2025. URL https://arxiv.org/abs/2502.
14866v2.
[1237] Shanglong Yang, Zhipeng Yuan, Shunbao Li, Ruoling Peng, Kang Liu, and Po Yang. Gpt-4 as evaluator:
Evaluating large language models on pest management in agriculture. arXiv preprint, 2024.
[1238] Wang Yang, Zirui Liu, Hongye Jin, Qingyu Yin, Vipin Chaudhary, and Xiaotian Han. Longer con-
text, deeper thinking: Uncovering the role of long-context ability in reasoning, arXiv preprint
arXiv:2505.17315, 2025. URL https://arxiv.org/abs/2505.17315v1.
[1239] Wen Yang, Kai Fan, and Minpeng Liao. Markov chain of thought for efficient mathematical reasoning.
North American Chapter of the Association for Computational Linguistics, 2024.
[1240] Yaodong Yang, Chengdong Ma, Zihan Ding, S. McAleer, Chi Jin, and Jun Wang. Game-theoretic
multiagent reinforcement learning, arXiv preprint arXiv:2011.00583, 2020. URL https://arxiv.
org/abs/2011.00583v4.
[1241] Yazheng Yang, Yuqi Wang, Sankalok Sen, Lei Li, and Qi Liu. Unleashing the potential of large
language models for predictive tabular tasks in data science, arXiv preprint arXiv:2403.20208, 2024.
URL https://arxiv.org/abs/2403.20208v7.
[1242] Yi Yang, Yixuan Tang, and Kar Yan Tam. Investlm: A large language model for investment using
financial domain instruction tuning, arXiv preprint arXiv:2309.13064, 2023. URL https://arxiv.
org/abs/2309.13064.
[1243] Yiben Yang, Chaitanya Malaviya, Jared Fernandez, Swabha Swayamdipta, Ronan Le Bras, Ji ping
Wang, Chandra Bhagavatula, Yejin Choi, and Doug Downey. G-daug: Generative data augmentation
for commonsense reasoning. Findings, 2020.
[1244] Yingxuan Yang, Huacan Chai, Yuanyi Song, Siyuan Qi, Muning Wen, Ning Li, Junwei Liao, Haoyi
Hu, Jianghao Lin, Gaowei Chang, Weiwen Liu, Ying Wen, Yong Yu, and Weinan Zhang. A survey
of ai agent protocols, arXiv preprint arXiv:2504.16736, 2025. URL https://arxiv.org/abs/
2504.16736v3.
[1245] Yuan Yang, Siheng Xiong, Ehsan Shareghi, and F. Fekri. The compressor-retriever architecture for
language model os, arXiv preprint arXiv:2409.01495, 2024. URL https://arxiv.org/abs/
2409.01495v1.
[1246] Yuxin Yang, Haoyang Wu, Tao Wang, Jia Yang, Hao Ma, and Guojie Luo. Pseudo-knowledge graph:
Meta-path guided retrieval and in-graph text for rag-equipped llm, arXiv preprint arXiv:2503.00309,
2025. URL https://arxiv.org/abs/2503.00309v1.
[1247] Zhen Yang, Fang Liu, Zhongxing Yu, J. Keung, Jia Li, Shuo Liu, Yifan Hong, Xiaoxue Ma, Zhi Jin, and
Ge Li. Exploring and unleashing the power of large language models in automated code translation.
Proc. ACM Softw. Eng., 2024.
[1248] Chengyuan Yao and Satoshi Fujita. Adaptive control of retrieval-augmented generation for large
language models through reflective tags. Electronics, 2024.
[1249] Huaiyuan Yao, Longchao Da, Vishnu Nandam, J. Turnau, Zhiwei Liu, Linsey Pang, and Hua Wei.
Comal: Collaborative multi-agent large language models for mixed-autonomy traffic, arXiv preprint
arXiv:2410.14368, 2024. URL https://arxiv.org/abs/2410.14368v2.
[1250] Jiayu Yao, Shenghua Liu, Yiwei Wang, Lingrui Mei, Baolong Bi, Yuyao Ge, Zhecheng Li, and Xueqi
Cheng. Who is in the spotlight: The hidden bias undermining multimodal retrieval-augmented
generation. 2025.
[1251] Jinghan Yao, Sam Ade Jacobs, Masahiro Tanaka, Olatunji Ruwase, A. Shafi, H. Subramoni, and
Dhabaleswar K. Panda. Training ultra long context language model with fully pipelined distributed
transformer, arXiv preprint arXiv:2408.16978, 2024. URL https://arxiv.org/abs/2408.
16978v2.
[1252] Liang Yao, Chengsheng Mao, and Yuan Luo. Graph convolutional networks for text classification.
AAAI Conference on Artificial Intelligence, 2018.
[1253] Shunyu Yao, Howard Chen, John Yang, and Karthik Narasimhan. Webshop: Towards scalable
real-world web interaction with grounded language agents. Neural Information Processing Systems,
2022.
[1254] Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, and Yuan Cao.
React: Synergizing reasoning and acting in language models. International Conference on Learning
Representations, 2022.
[1255] Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, T. Griffiths, Yuan Cao, and Karthik Narasimhan.
Tree of thoughts: Deliberate problem solving with large language models. Neural Information
Processing Systems, 2023.
[1256] Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, and Yuan Cao.
React: Synergizing reasoning and acting in language models, arXiv preprint arXiv:2210.03629, 2023.
URL https://arxiv.org/abs/2210.03629.
[1257] Shunyu Yao, Noah Shinn, P. Razavi, and Karthik Narasimhan. τ-bench: A benchmark for tool-agent-
user interaction in real-world domains. arXiv preprint, 2024.
[1258] Weiran Yao, Shelby Heinecke, Juan Carlos Niebles, Zhiwei Liu, Yihao Feng, Le Xue, Rithesh Murthy,
Zeyuan Chen, Jianguo Zhang, Devansh Arpit, Ran Xu, Phil Mui, Huan Wang, Caiming Xiong, and
Silvio Savarese. Retroformer: Retrospective large language agents with policy gradient optimization,
arXiv preprint arXiv:2308.02151, 2024. URL https://arxiv.org/abs/2308.02151.
[1259] Michihiro Yasunaga, Hongyu Ren, Antoine Bosselut, Percy Liang, and J. Leskovec. Qa-gnn: Reasoning
with language models and knowledge graphs for question answering. North American Chapter of the
Association for Computational Linguistics, 2021.
[1260] Michihiro Yasunaga, Antoine Bosselut, Hongyu Ren, Xikun Zhang, Christopher D. Manning, Percy
Liang, and J. Leskovec. Deep bidirectional language-knowledge graph pretraining. Neural Information
Processing Systems, 2022.
[1261] Fahd Yazin, Moumita Das, A. Banerjee, and Dipanjan Roy. Contextual prediction errors reorganize
naturalistic episodic memories in time. Scientific Reports, 2021.
[1262] J Ye, G Li, S Gao, C Huang, Y Wu, S Li, and X Fan. . . . Tooleyes: Fine-grained evaluation for
tool learning capabilities of large language models in real-world scenarios. 2024. URL https:
//arxiv.org/abs/2401.00741.
[1263] J Ye, S Li, G Li, C Huang, S Gao, and Y Wu. . . . Toolsword: Unveiling safety issues of large language
models in tool learning across three stages. 2024. URL https://arxiv.org/abs/2402.10753.
[1264] Junjie Ye, Zhengyin Du, Xuesong Yao, Weijian Lin, Yufei Xu, Zehui Chen, Zaiyuan Wang, Sining
Zhu, Zhiheng Xi, Siyu Yuan, Tao Gui, Qi Zhang, Xuanjing Huang, and Jiechao Chen. Toolhop: A
query-driven benchmark for evaluating large language models in multi-hop tool use, arXiv preprint
arXiv:2501.02506, 2025. URL https://arxiv.org/abs/2501.02506v4.
[1265] Qinyuan Ye, Maxamed Axmed, Reid Pryzant, and Fereshte Khani. Prompt engineering a prompt
engineer. Annual Meeting of the Association for Computational Linguistics, 2023.
[1266] Sixiang Ye, Zeyu Sun, Guoqing Wang, Liwei Guo, Qing-Lin Liang, Zheng Li, and Yong Liu.
Prompt alchemy: Automatic prompt refinement for enhancing code generation, arXiv preprint
arXiv:2503.11085, 2025. URL https://arxiv.org/abs/2503.11085v1.
[1267] Zhifan Ye, Kejing Xia, Yonggan Fu, Xin Dong, Jihoon Hong, Xiangchi Yuan, Shizhe Diao, Jan
Kautz, Pavlo Molchanov, and Y. Lin. Longmamba: Enhancing mamba’s long-context capabilities
via training-free receptive field enlargement. International Conference on Learning Representations,
2025.
[1268] Asaf Yehudai, Lilach Eden, Alan Li, Guy Uziel, Yilun Zhao, Roy Bar-Haim, Arman Cohan, and Michal
Shmueli-Scheuer. Survey on evaluation of llm-based agents, arXiv preprint arXiv:2503.16416, 2025.
URL https://arxiv.org/abs/2503.16416v1.
[1269] Peiling Yi and Yuhan Xia. Irony detection, reasoning and understanding in zero-shot learning. IEEE
Transactions on Artificial Intelligence, 2025.
[1270] Da Yin, Faeze Brahman, Abhilasha Ravichander, Khyathi Raghavi Chandu, Kai-Wei Chang, Yejin Choi,
and Bill Yuchen Lin. Agent lumos: Unified and modular training for open-source language agents.
Annual Meeting of the Association for Computational Linguistics, 2023.
[1271] Fan Yin, Zifeng Wang, I-Hung Hsu, Jun Yan, Ke Jiang, Yanfei Chen, Jindong Gu, Long T. Le,
Kai-Wei Chang, Chen-Yu Lee, Hamid Palangi, and Tomas Pfister. Magnet: Multi-turn tool-use
data synthesis and distillation via graph translation, arXiv preprint arXiv:2503.07826, 2025. URL
https://arxiv.org/abs/2503.07826v1.
[1272] Guoli Yin, Haoping Bai, Shuang Ma, Feng Nan, Yanchao Sun, Zhaoyang Xu, Shen Ma, Jiarui Lu,
Xiang Kong, Aonan Zhang, Dian Ang Yap, Yizhe Zhang, K. Ahnert, Vik Kamath, Mathias Berglund,
Dominic Walsh, Tobias Gindele, Juergen Wiest, Zhengfeng Lai, Xiaoming Wang, Jiulong Shan, Meng
Cao, Ruoming Pang, and Zirui Wang. Mmau: A holistic benchmark of agent capabilities across
diverse domains. North American Chapter of the Association for Computational Linguistics, 2024.
[1273] Pengcheng Yin, Graham Neubig, Wen tau Yih, and Sebastian Riedel. Tabert: Pretraining for joint
understanding of textual and tabular data. Annual Meeting of the Association for Computational
Linguistics, 2020.
[1274] S Yin, W You, Z Ji, G Zhong, and J Bai. Mumath-code: Combining tool-use large language models
with multi-perspective data augmentation for mathematical reasoning. 2024. URL https://
arxiv.org/abs/2405.07551.
[1275] Gunwoo Yong, Kahyun Jeon, Daeyoung Gil, and Ghang Lee. Prompt engineering for zero-shot and
few-shot defect detection and classification using a visual-language pretrained model. Comput. Aided
Civ. Infrastructure Eng., 2022.
[1276] Aspen H. Yoo and A. Collins. How working memory and reinforcement learning are intertwined: A
cognitive, neural, and computational perspective. Journal of Cognitive Neuroscience, 2021.
[1277] Chanwoong Yoon, Taewhoo Lee, Hyeon Hwang, Minbyul Jeong, and Jaewoo Kang. Compact:
Compressing retrieved documents actively for question answering. Conference on Empirical Methods
in Natural Language Processing, 2024.
[1278] Jiaxuan You, Mingjie Liu, Shrimai Prabhumoye, M. Patwary, M. Shoeybi, and Bryan Catanzaro.
Llm-evolve: Evaluation for llm’s evolving capability on benchmarks. Conference on Empirical Methods
in Natural Language Processing, 2024.
[1279] Yuxin You, Zhen Liu, Xiangchao Wen, Yongtao Zhang, and Wei Ai. Large language models meet
graph neural networks: A perspective of graph mining. Mathematics, 2024.
[1280] Dian Yu, Yuheng Zhang, Jiahao Xu, Tian Liang, Linfeng Song, Zhaopeng Tu, Haitao Mi, and
Dong Yu. Teaching llms to refine with tools, arXiv preprint arXiv:2412.16871, 2024. URL https:
//arxiv.org/abs/2412.16871v1.
[1281] Miao Yu, Fanci Meng, Xinyun Zhou, Shilong Wang, Junyuan Mao, Linsey Pang, Tianlong Chen,
Kun Wang, Xinfeng Li, Yongfeng Zhang, Bo An, and Qingsong Wen. A survey on trustworthy
llm agents: Threats and countermeasures, arXiv preprint arXiv:2503.09648, 2025. URL https:
//arxiv.org/abs/2503.09648v1.
[1282] Ye Yu, Yaoning Yu, and Haohan Wang. Premise: Scalable and strategic prompt optimization for
efficient mathematical reasoning in large models, arXiv preprint arXiv:2506.10716, 2025. URL
https://arxiv.org/abs/2506.10716v1.
[1283] Zeping Yu and Sophia Ananiadou. Understanding multimodal llms: the mechanistic interpretability
of llava in visual question answering, arXiv preprint arXiv:2411.10950v2, 2024. URL https:
//arxiv.org/abs/2411.10950v2.
[1284] Zishun Yu, Tengyu Xu, Di Jin, Karthik Abinav Sankararaman, Yun He, Wenxuan Zhou, Zhouhao
Zeng, Eryk Helenowski, Chen Zhu, Si-Yuan Wang, Hao Ma, and Han Fang. Think smarter not harder:
Adaptive reasoning with inference aware optimization, arXiv preprint arXiv:2501.17974, 2025. URL
https://arxiv.org/abs/2501.17974v2.
[1285] Zhao yu Su, Linjie Li, Mingyang Song, Yunzhuo Hao, Zhengyuan Yang, Jun Zhang, Guanjie Chen,
Jiawei Gu, Juntao Li, Xiaoye Qu, and Yu Cheng. Openthinkimg: Learning to think with images via
visual tool reinforcement learning, arXiv preprint arXiv:2505.08617, 2025. URL https://arxiv.
org/abs/2505.08617v1.
[1286] Siyu Yuan, Zehui Chen, Zhiheng Xi, Junjie Ye, Zhengyin Du, and Jiecao Chen. Agent-r: Training
language model agents to reflect via iterative self-training. arXiv preprint, 2025.
[1287] Weizhe Yuan, Richard Yuanzhe Pang, Kyunghyun Cho, Sainbayar Sukhbaatar, Jing Xu, and Jason E
Weston. Self-rewarding language models. International Conference on Machine Learning, 2024.
[1288] Xiaowei Yuan, Zhao Yang, Ziyang Huang, Yequan Wang, Siqi Fan, Yiming Ju, Jun Zhao, and Kang
Liu. Exploiting contextual knowledge in llms through v-usable information based layer enhancement.
arXiv preprint, 2025.
[1289] Xinbin Yuan, Jian Zhang, Kaixin Li, Zhuoxuan Cai, Lujian Yao, Jie Chen, Enguang Wang, Qibin
Hou, Jinwei Chen, Peng-Tao Jiang, and Bo Li. Enhancing visual grounding for gui agents via
self-evolutionary reinforcement learning, arXiv preprint arXiv:2505.12370, 2025. URL https:
//arxiv.org/abs/2505.12370v2.
[1290] Murong Yue. A survey of large language model agents for question answering, arXiv preprint
arXiv:2503.19213, 2025. URL https://arxiv.org/abs/2503.19213v1.
[1291] Xihang Yue, Linchao Zhu, and Yi Yang. Fragrel: Exploiting fragment-level relations in the external
memory of large language models. Annual Meeting of the Association for Computational Linguistics,
2024.
[1292] Seongjun Yun, Minbyul Jeong, Raehyun Kim, Jaewoo Kang, and Hyunwoo J. Kim. Graph transformer
networks. Neural Information Processing Systems, 2019.
[1293] Ge Yuyao, Cheng Yiting, Wang Jia, Zhou Hanlin, and Chen Lizhe. Vision transformer based on
knowledge distillation in tcm image classification. In 2022 IEEE 5th International Conference on
Computer and Communication Engineering Technology (CCET), pages 120–125. IEEE, 2022.
[1294] M. Zaheer, Guru Guruganesh, Kumar Avinava Dubey, J. Ainslie, Chris Alberti, Santiago Ontañón,
Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, and Amr Ahmed. Big bird: Transformers for
longer sequences. Neural Information Processing Systems, 2020.
[1295] Yuhang Zang, Wei Li, Jun Han, Kaiyang Zhou, and Chen Change Loy. Contextual object detection
with multimodal large language models. International Journal of Computer Vision, 2023.
[1296] E. Zelikman, Yuhuai Wu, and Noah D. Goodman. Star: Bootstrapping reasoning with reasoning,
arXiv preprint arXiv:2203.14465, 2022. URL https://arxiv.org/abs/2203.14465v2.
[1297] E. Zelikman, Eliana Lorch, Lester Mackey, and A. Kalai. Self-taught optimizer (stop): Recursively
self-improving code generation, arXiv preprint arXiv:2310.02304, 2023. URL https://arxiv.
org/abs/2310.02304v3.
[1298] Pai Zeng, Zhenyu Ning, Jieru Zhao, Weihao Cui, Mengwei Xu, Liwei Guo, XuSheng Chen, and Yizhou
Shan. The cap principle for llm serving: A survey of long-context large language model serving,
arXiv preprint arXiv:2405.11299, 2024. URL https://arxiv.org/abs/2405.11299v2.
[1299] Ruihong Zeng, Jinyuan Fang, Siwei Liu, and Zaiqiao Meng. On the structural memory of llm agents,
arXiv preprint arXiv:2412.15266, 2024. URL https://arxiv.org/abs/2412.15266v1.
[1300] Yirong Zeng, Xiao Ding, Yuxian Wang, Weiwen Liu, Wu Ning, Yutai Hou, Xu Huang, Bing Qin, and
Ting Liu. itool: Reinforced fine-tuning with dynamic deficiency calibration for advanced tool use,
arXiv preprint arXiv:2501.09766, 2025. URL https://arxiv.org/abs/2501.09766v4.
[1301] Yongcheng Zeng, Xinyu Cui, Xuanfa Jin, Guoqing Liu, Zexu Sun, Dong Li, Ning Yang, Jianye Hao,
Haifeng Zhang, and Jun Wang. Evolving llms’ self-refinement capability via iterative preference
optimization, arXiv preprint arXiv:2502.05605, 2025. URL https://arxiv.org/abs/2502.
05605v3.
[1302] An Zhang, Leheng Sheng, Yuxin Chen, Hao Li, Yang Deng, Xiang Wang, and Tat-Seng Chua. On
generative agents in recommendation. Annual International ACM SIGIR Conference on Research and
Development in Information Retrieval, 2023.
[1303] B Zhang, K Zhou, X Wei, and X Zhao. . . . Evaluating and improving tool-augmented computation-
intensive math reasoning. 2023. URL https://proceedings.neurips.cc/paper_files/
paper/2023/hash/4a47dd69242d5af908cdd5d51c971cbf-Abstract-Datasets_and_
Benchmarks.html.
[1304] Chao Zhang, Haoxin Zhang, Shiwei Wu, Di Wu, Tong Xu, Xiangyu Zhao, Yan Gao, Yao Hu, and
Enhong Chen. Notellm-2: Multimodal large representation models for recommendation. Knowledge
Discovery and Data Mining, 2024.
[1305] Chaoyun Zhang, He Huang, Chiming Ni, Jian Mu, Si Qin, Shilin He, Lu Wang, Fangkai Yang, Pu Zhao,
Chao Du, et al. Ufo2: The desktop agentos. arXiv preprint arXiv:2504.14603, 2025.
[1306] Chi Zhang, Yujun Cai, Guosheng Lin, and Chunhua Shen. Deepemd: Few-shot image classification
with differentiable earth mover’s distance and structured classifiers. In Proceedings of the IEEE/CVF
conference on computer vision and pattern recognition, pages 12203–12213, 2020.
[1307] Dan Zhang, G. Feng, Yang Shi, and D. Srinivasan. Physical safety and cyber security analysis of
multi-agent systems: A survey of recent advances. IEEE/CAA Journal of Automatica Sinica, 2021.
[1308] Danyang Zhang, Lu Chen, Situo Zhang, Hongshen Xu, Zihan Zhao, and Kai Yu. Large language
models are semi-parametric reinforcement learning agents. Neural Information Processing Systems,
2023.
[1309] Daoan Zhang, Weitong Zhang, Bing He, Jiang Zhang, Chenchen Qin, and Jianhua Yao. Dnagpt: A
generalized pre-trained tool for multiple dna sequence analysis tasks. bioRxiv, 2024.
[1310] Duzhen Zhang, Yahan Yu, Jiahua Dong, Chenxing Li, Dan Su, Chenhui Chu, and Dong Yu. Mm-
llms: Recent advances in multimodal large language models. In Findings of the Association for
Computational Linguistics ACL 2024, pages 12401–12430, 2024.
[1311] Duzhen Zhang, Yong Ren, Zhong-Zhi Li, Yahan Yu, Jiahua Dong, Chenxing Li, Zhilong Ji, and Jinfeng
Bai. Enhancing multimodal continual instruction tuning with branchlora. In Proceedings of the 63rd
Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2025.
[1312] Han Zhang, Langshi Zhou, and Hanfang Yang. Learning to retrieve and reason on knowledge graph
through active self-reflection, arXiv preprint arXiv:2502.14932, 2025. URL https://arxiv.org/
abs/2502.14932v1.
[1313] Hengyu Zhang. Sinklora: Enhanced efficiency and chat capabilities for long-context large lan-
guage models, arXiv preprint arXiv:2406.05678, 2024. URL https://arxiv.org/abs/2406.
05678v1.
[1314] Jiaxin Zhang, Zhongzhi Li, Mingliang Zhang, Fei Yin, Chenglin Liu, and Yashar Moshfeghi. Geoeval:
benchmark for evaluating llms and multi-modal models on geometry problem-solving. 2024.
[1315] Jing Zhang, Xiaokang Zhang, Jifan Yu, Jian Tang, Jie Tang, Cuiping Li, and Hong Chen. Subgraph
retrieval enhanced model for multi-hop knowledge base question answering. Annual Meeting of the
Association for Computational Linguistics, 2022.
[1316] Kai Zhang, Fubang Zhao, Yangyang Kang, and Xiaozhong Liu. Llm-based medical assistant personal-
ization with short- and long-term memory coordination. North American Chapter of the Association
for Computational Linguistics, 2023.
[1317] Kai Zhang, Yejin Kim, and Xiaozhong Liu. Personalized llm response generation with parameterized
memory injection, arXiv preprint arXiv:2404.03565, 2025. URL https://arxiv.org/abs/2404.
03565.
[1318] Kechi Zhang, Zhuo Li, Jia Li, Ge Li, and Zhi Jin. Self-edit: Fault-aware code editor for code generation.
Annual Meeting of the Association for Computational Linguistics, 2023.
[1319] Kechi Zhang, Jia Li, Ge Li, Xianjie Shi, and Zhi Jin. Codeagent: Enhancing code generation with
tool-integrated agent systems for real-world repo-level coding challenges. Annual Meeting of the
Association for Computational Linguistics, 2024.
[1320] Kechi Zhang, Ge Li, Jia Li, Huangzhao Zhang, Jingjing Xu, Hao Zhu, Lecheng Wang, Yihong Dong,
Jing Mai, Bin Gu, and Zhi Jin. Computational thinking reasoning in large language models, arXiv
preprint arXiv:2506.02658, 2025. URL https://arxiv.org/abs/2506.02658v2.
[1321] Ming-Liang Zhang, Zhong-Zhi Li, Fei Yin, Liang Lin, and Cheng-Lin Liu. Fuse, reason and verify:
Geometry problem solving with parsed clauses from diagram. 2024.
[1322] Qiyuan Zhang, Fuyuan Lyu, Zexu Sun, Lei Wang, Weixu Zhang, Zhihan Guo, Yufei Wang, Irwin
King, Xue Liu, and Chen Ma. A survey on test-time scaling in large language models: What, how,
where, and how well?, arXiv preprint arXiv:2503.24235, 2025. URL https://arxiv.org/abs/
2503.24235v3.
[1323] Ruichen Zhang, Mufan Qiu, Zhen Tan, Mohan Zhang, Vincent Lu, Jie Peng, Kaidi Xu, Leandro Z.
Agudelo, Peter Qian, and Tianlong Chen. Symbiotic cooperation for web agents: Harnessing
complementary strengths of large and small llms, arXiv preprint arXiv:2502.07942, 2025. URL
https://arxiv.org/abs/2502.07942v2.
[1324] Tengchao Zhang, Yonglin Tian, Fei Lin, Jun Huang, Patrik P. Süli, Rui Qin, and Fei-Yue Wang.
Coordfield: Coordination field for agentic uav task allocation in low-altitude urban scenarios, arXiv
preprint arXiv:2505.00091, 2025. URL https://arxiv.org/abs/2505.00091v3.
[1325] Weizhi Zhang, Xinyang Zhang, Chenwei Zhang, Liangwei Yang, Jingbo Shang, Zhepei Wei,
Henry Peng Zou, Zijie Huang, Zhengyang Wang, Yifan Gao, Xiaoman Pan, Lian Xiong, Jingguo
Liu, Philip S. Yu, and Xian Li. Personaagent: When large language model agents meet personaliza-
tion at test time, arXiv preprint arXiv:2506.06254, 2025. URL https://arxiv.org/abs/2506.
06254v1.
[1326] Wen Zhang, Long Jin, Yushan Zhu, Jiaoyan Chen, Zhiwei Huang, Junjie Wang, Yin Hua, Lei Liang,
and Hua zeng Chen. Trustuqa: A trustful framework for unified structured data question answering.
AAAI Conference on Artificial Intelligence, 2024.
[1327] Wenlin Zhang, Xiangyang Li, Kuicai Dong, Yichao Wang, Pengyue Jia, Xiaopeng Li, Yingyi Zhang,
Derong Xu, Zhaochen Du, Huifeng Guo, Ruiming Tang, and Xiangyu Zhao. Process vs. outcome
reward: Which is better for agentic rag reinforcement learning, arXiv preprint arXiv:2505.14069,
2025. URL https://arxiv.org/abs/2505.14069v2.
[1328] Wentao Zhang, Ce Cui, Yilei Zhao, Rui Hu, Yang Liu, Yahui Zhou, and Bo An. Agentorchestra: A hi-
erarchical multi-agent framework for general-purpose task solving, arXiv preprint arXiv:2506.12508,
2025. URL https://arxiv.org/abs/2506.12508v2.
[1329] Xiaoyi Zhang, Zhaoyang Jia, Zongyu Guo, Jiahao Li, Bin Li, Houqiang Li, and Yan Lu. Deep
video discovery: Agentic search with tool use for long-form video understanding, arXiv preprint
arXiv:2505.18079, 2025. URL https://arxiv.org/abs/2505.18079v2.
[1330] Xikun Zhang, Antoine Bosselut, Michihiro Yasunaga, Hongyu Ren, Percy Liang, Christopher D.
Manning, and J. Leskovec. Greaselm: Graph reasoning enhanced language models for question
answering. International Conference on Learning Representations, 2022.
[1331] Yao Zhang, Zijian Ma, Yunpu Ma, Zhen Han, Yu Wu, and Volker Tresp. Webpilot: A versatile and
autonomous multi-agent system for web task execution with strategic exploration, arXiv preprint
arXiv:2408.15978, 2024. URL https://arxiv.org/abs/2408.15978.
[1332] Yinger Zhang, Hui Cai, Yicheng Chen, Rui Sun, and Jing Zheng. Reverse chain: A generic-rule for
llms to master multi-api planning, arXiv preprint arXiv:2310.04474, 2023. URL https://arxiv.
org/abs/2310.04474v3.
[1333] Yongheng Zhang, Qiguang Chen, Jingxuan Zhou, Peng Wang, Jiasheng Si, Jin Wang, Wenpeng
Lu, and Libo Qin. Wrong-of-thought: An integrated reasoning framework with multi-perspective
verification and wrong information, arXiv preprint arXiv:2410.04463, 2024. URL https://arxiv.
org/abs/2410.04463.
[1334] Youjia Zhang, Jin Wang, Liang-Chih Yu, and Xuejie Zhang. Ma-bert: Learning representation by
incorporating multi-attribute knowledge in transformers. Findings, 2021.
[1335] Yu Zhang, Jinlong Ma, Yongshuai Hou, Xuefeng Bai, Kehai Chen, Yang Xiang, Jun Yu, and Min
Zhang. Evaluating and steering modality preferences in multimodal large language model, arXiv
preprint arXiv:2505.20977v1, 2025. URL https://arxiv.org/abs/2505.20977v1.
[1336] Yunyi Zhang, Ming Zhong, Siru Ouyang, Yizhu Jiao, Sizhe Zhou, Linyi Ding, and Jiawei Han.
Automated mining of structured knowledge from text in the era of large language models. Knowledge
Discovery and Data Mining, 2024.
[1337] Yusen Zhang, Ruoxi Sun, Yanfei Chen, Tomas Pfister, Rui Zhang, and Sercan Ö. Arik. Chain of agents:
Large language models collaborating on long-context tasks. Neural Information Processing Systems,
2024.
[1338] Yuxiang Zhang, Yuqi Yang, Jiangming Shu, Xinyan Wen, and Jitao Sang. Agent models: Internalizing
chain-of-action generation into reasoning models, arXiv preprint arXiv:2503.06580, 2025. URL
https://arxiv.org/abs/2503.06580v1.
[1339] Zeyu Zhang, Xiaohe Bo, Chen Ma, Rui Li, Xu Chen, Quanyu Dai, Jieming Zhu, Zhenhua Dong, and
Ji-Rong Wen. A survey on the memory mechanism of large language model based agents, arXiv
preprint arXiv:2404.13501, 2024. URL https://arxiv.org/abs/2404.13501v1.
[1340] Zeyu Zhang, Quanyu Dai, Luyu Chen, Zeren Jiang, Rui Li, Jieming Zhu, Xu Chen, Yi Xie, Zhenhua
Dong, and Ji-Rong Wen. Memsim: A bayesian simulator for evaluating memory of llm-based
personal assistants, arXiv preprint arXiv:2409.20163, 2024. URL https://arxiv.org/abs/
2409.20163v1.
[1341] Zeyu Zhang, Quanyu Dai, Xu Chen, Rui Li, Zhongyang Li, and Zhenhua Dong. Memengine: A unified
and modular library for developing advanced memory of llm-based agents. The Web Conference,
2025.
[1342] Zheng Zhang, Liang Ding, Dazhao Cheng, Xuebo Liu, Min Zhang, and Dacheng Tao. Bliss:
Robust sequence-to-sequence learning via self-supervised input representation, arXiv preprint
arXiv:2204.07837, 2022. URL https://arxiv.org/abs/2204.07837v2.
[1343] Zhenyu (Allen) Zhang, Ying Sheng, Tianyi Zhou, Tianlong Chen, Lianmin Zheng, Ruisi Cai, Zhao
Song, Yuandong Tian, Christopher Ré, Clark W. Barrett, Zhangyang Wang, and Beidi Chen. H2o:
Heavy-hitter oracle for efficient generative inference of large language models. Neural Information
Processing Systems, 2023.
[1344] Zhihan Zhang, Zhenwen Liang, Wenhao Yu, Dian Yu, Mengzhao Jia, Dong Yu, and Meng Jiang.
Learn beyond the answer: Training language models with reflection for mathematical reasoning.
Conference on Empirical Methods in Natural Language Processing, 2024.
[1345] Zhuosheng Zhang and Aston Zhang. You only look at screens: Multimodal chain-of-action agents.
Annual Meeting of the Association for Computational Linguistics, 2023.
[1346] Andrew Zhao, Daniel Huang, Quentin Xu, Matthieu Lin, Y. Liu, and Gao Huang. Expel: Llm agents
are experiential learners. AAAI Conference on Artificial Intelligence, 2023.
[1347] Andrew Zhao, Daniel Huang, Quentin Xu, Matthieu Lin, Yong-Jin Liu, and Gao Huang. Expel: Llm
agents are experiential learners, arXiv preprint arXiv:2308.10144, 2024. URL https://arxiv.
org/abs/2308.10144.
[1348] Jianan Zhao, Meng Qu, Chaozhuo Li, Hao Yan, Qian Liu, Rui Li, Xing Xie, and Jian Tang. Learning
on large-scale text-attributed graphs via variational inference. International Conference on Learning
Representations, 2022.
[1349] Penghao Zhao, Hailin Zhang, Qinhan Yu, Zhengren Wang, Yunteng Geng, Fangcheng Fu, Ling Yang,
Wentao Zhang, and Bin Cui. Retrieval-augmented generation for ai-generated content: A survey,
arXiv preprint arXiv:2402.19473, 2024. URL https://arxiv.org/abs/2402.19473v6.
[1350] Pengyu Zhao, Zijian Jin, and Ning Cheng. An in-depth survey of large language model-based artificial
intelligence agents, arXiv preprint arXiv:2309.14365, 2023. URL https://arxiv.org/abs/
2309.14365v1.
[1351] Pu Zhao, Xuan Shen, Zhenglun Kong, Yixin Shen, Sung-En Chang, Timothy Rupprecht, Lei Lu, Enfu
Nan, Changdi Yang, Yumei He, Xingchen Xu, Yu Huang, Wei Wang, Yue Chen, Yongchun He, and
Yanzhi Wang. 7b fully open source moxin-llm/vlm – from pretraining to grpo-based reinforcement
learning enhancement. arXiv preprint, 2024.
[1352] Qi Zhao, Hongyu Yang, Qi Song, Xinwei Yao, and Xiangyang Li. Knowpath: Knowledge-enhanced
reasoning via llm-generated inference paths over knowledge graphs, arXiv preprint arXiv:2502.12029,
2025. URL https://arxiv.org/abs/2502.12029v3.
[1353] Qifang Zhao, Weidong Ren, Tianyu Li, Xiaoxiao Xu, and Hong Liu. Graphgpt: Generative pre-trained
graph eulerian transformer, arXiv preprint arXiv:2401.00529v3, 2023. URL https://arxiv.org/
abs/2401.00529v3.
[1354] Ruilin Zhao, Feng Zhao, Long Wang, Xianzhi Wang, and Guandong Xu. Kg-cot: Chain-of-thought
prompting of large language models over knowledge graphs for knowledge-aware question answering.
International Joint Conference on Artificial Intelligence, 2024.
[1355] Shangziqi Zhao, Jiahao Yuan, Guisong Yang, and Usman Naseem. Can pruning improve reason-
ing? revisiting long-cot compression with capability in mind for better reasoning, arXiv preprint
arXiv:2505.14582, 2025. URL https://arxiv.org/abs/2505.14582v1.
[1356] Shitian Zhao, Zhuowan Li, Yadong Lu, Alan L. Yuille, and Yan Wang. Causal-cog: A causal-effect
look at context generation for boosting multi-modal language models. Computer Vision and Pattern
Recognition, 2023.
[1357] Tony Zhao, Eric Wallace, Shi Feng, D. Klein, and Sameer Singh. Calibrate before use: Improving
few-shot performance of language models. International Conference on Machine Learning, 2021.
[1358] Weixiang Zhao, Xingyu Sui, Jiahe Guo, Yulin Hu, Yang Deng, Yanyan Zhao, Bing Qin, Wanxiang
Che, Tat-Seng Chua, and Ting Liu. Trade-offs in large reasoning models: An empirical analysis of
deliberative and adaptive reasoning over foundational capabilities, arXiv preprint arXiv:2503.17979,
2025. URL https://arxiv.org/abs/2503.17979v1.
[1359] Yibo Zhao, Jiapeng Zhu, Ye Guo, Kangkang He, and Xiang Li. E2 graphrag: Streamlining graph-based
rag for high efficiency and effectiveness. arXiv preprint, 2025.
[1360] Boyuan Zheng, Boyu Gou, Jihyung Kil, Huan Sun, and Yu Su. Gpt-4v(ision) is a generalist web
agent, if grounded. International Conference on Machine Learning, 2024.
[1361] Changmeng Zheng, Dayong Liang, Wengyu Zhang, Xiao Wei, Tat seng Chua, and Qing Li. A picture
is worth a graph: A blueprint debate paradigm for multimodal reasoning. ACM Multimedia, 2024.
[1362] Chuanyang Zheng, Yihang Gao, Han Shi, Minbin Huang, Jingyao Li, Jing Xiong, Xiaozhe Ren,
Michael Ng, Xin Jiang, Zhenguo Li, and Yu Li. Dape: Data-adaptive positional encoding for length
extrapolation. Neural Information Processing Systems, 2024.
[1363] Chunmo Zheng, Saika Wong, Xing Su, Yinqiu Tang, Ahsan Nawaz, and Mohamad Kassem. Automating
construction contract review using knowledge graph-enhanced large language models. Automation
in Construction, 2023.
[1364] Junhao Zheng, Shengjie Qiu, Chengming Shi, and Qianli Ma. Towards lifelong learning of large
language models: A survey. ACM Computing Surveys, 2024.
[1365] Junhao Zheng, Xidi Cai, Qiuke Li, Duzhen Zhang, Zhongzhi Li, Yingying Zhang, Le Song, and Qianli
Ma. Lifelongagentbench: Evaluating llm agents as lifelong learners, arXiv preprint arXiv:2505.11942,
2025. URL https://arxiv.org/abs/2505.11942v3.
[1366] Longtao Zheng, R. Wang, and Bo An. Synapse: Trajectory-as-exemplar prompting with memory for
computer control. International Conference on Learning Representations, 2023.
[1367] Longtao Zheng, Rundong Wang, Xinrun Wang, and Bo An. Synapse: Trajectory-as-exemplar
prompting with memory for computer control, arXiv preprint arXiv:2306.07863, 2024. URL
https://arxiv.org/abs/2306.07863.
[1368] Xu Zheng, Chenfei Liao, Yuqian Fu, Kaiyu Lei, Yuanhuiyi Lyu, Lutao Jiang, Bin Ren, Jialei Chen,
Jiawen Wang, Chengxin Li, Linfeng Zhang, D. Paudel, Xuanjing Huang, Yu-Gang Jiang, N. Sebe,
Dacheng Tao, L. V. Gool, and Xuming Hu. Mllms are deeply affected by modality bias, arXiv preprint
arXiv:2505.18657v1, 2025. URL https://arxiv.org/abs/2505.18657v1.
[1369] Yuxiang Zheng, Dayuan Fu, Xiangkun Hu, Xiaojie Cai, Lyumanshan Ye, Pengrui Lu, and Pengfei Liu.
Deepresearcher: Scaling deep research via reinforcement learning in real-world environments, arXiv
preprint arXiv:2504.03160, 2025. URL https://arxiv.org/abs/2504.03160v4.
[1370] Li Zhong, Zilong Wang, and Jingbo Shang. Debug like a human: A large language model debugger
via verifying runtime execution step by step. Annual Meeting of the Association for Computational
Linguistics, 2024.
[1371] Rui Zhong, Yang Cao, Jun Yu, and M. Munetomo. Large language model assisted adversarial
robustness neural architecture search. 2024 6th International Conference on Data-driven Optimization
of Complex Systems (DOCS), 2024.
[1372] Wanjun Zhong, Lianghong Guo, Qi-Fei Gao, He Ye, and Yanlin Wang. Memorybank: Enhancing
large language models with long-term memory. AAAI Conference on Artificial Intelligence, 2023.
[1373] Wanjun Zhong, Lianghong Guo, Qiqi Gao, He Ye, and Yanlin Wang. Memorybank: Enhancing
large language models with long-term memory, arXiv preprint arXiv:2305.10250, 2023. URL
https://arxiv.org/abs/2305.10250.
[1374] Andy Zhou, Kai Yan, Michal Shlapentokh-Rothman, Haohan Wang, and Yu-Xiong Wang. Language
agent tree search unifies reasoning acting and planning in language models. International Conference
on Machine Learning, 2023.
[1375] Bin Zhou, Xingwang Shen, Yuqian Lu, Xinyu Li, B. Hua, Tianyuan Liu, and Jinsong Bao. Semantic-
aware event link reasoning over industrial knowledge graph embedding time series data. International
Journal of Production Research, 2022.
[1376] Denny Zhou, Nathanael Scharli, Le Hou, Jason Wei, Nathan Scales, Xuezhi Wang, D. Schuurmans,
O. Bousquet, Quoc Le, and Ed H. Chi. Least-to-most prompting enables complex reasoning in large
language models. International Conference on Learning Representations, 2022.
[1377] Huichi Zhou, Kin-Hei Lee, Zhonghao Zhan, Yue Chen, Zhenhao Li, Zhaoyang Wang, Hamed Haddadi,
and Emine Yilmaz. Trustrag: Enhancing robustness and trustworthiness in rag, arXiv preprint
arXiv:2501.00879, 2025. URL https://arxiv.org/abs/2501.00879.
[1378] Shuyan Zhou, Frank F. Xu, Hao Zhu, Xuhui Zhou, Robert Lo, Abishek Sridhar, Xianyi Cheng, Yonatan
Bisk, Daniel Fried, Uri Alon, and Graham Neubig. Webarena: A realistic web environment for
building autonomous agents. International Conference on Learning Representations, 2023.
[1379] Wangchunshu Zhou, Y. Jiang, Long Li, Jialong Wu, Tiannan Wang, Shi Qiu, Jintian Zhang, Jing
Chen, Ruipu Wu, Shuai Wang, Shiding Zhu, Jiyu Chen, Wentao Zhang, Xiangru Tang, Ningyu Zhang,
Huajun Chen, Peng Cui, and Mrinmaya Sachan. Agents: An open-source framework for autonomous
language agents, arXiv preprint arXiv:2309.07870, 2023. URL https://arxiv.org/abs/2309.
07870v3.
[1380] Yingli Zhou, Yaodong Su, Youran Sun, Shu Wang, Taotao Wang, Runyuan He, Yongwei Zhang,
Sicong Liang, Xilin Liu, Yuchi Ma, and Yixiang Fang. In-depth analysis of graph-based rag in a
unified framework, arXiv preprint arXiv:2503.04338, 2025. URL https://arxiv.org/abs/
2503.04338v1.
[1381] Yuhang Zhou and Wei Ai. Teaching-assistant-in-the-loop: Improving knowledge distillation from
imperfect teacher models in low-budget scenarios. Annual Meeting of the Association for Computational
Linguistics, 2024.
[1382] Yujia Zhou, Yan Liu, Xiaoxi Li, Jiajie Jin, Hongjin Qian, Zheng Liu, Chaozhuo Li, Zhicheng Dou,
Tsung-Yi Ho, and Philip S. Yu. Trustworthiness in retrieval-augmented generation systems: A survey,
arXiv preprint arXiv:2409.10102, 2024. URL https://arxiv.org/abs/2409.10102v1.
[1383] Zhehua Zhou, Jiayang Song, Kunpeng Yao, Zhan Shu, and Lei Ma. Isr-llm: Iterative self-refined
large language model for long-horizon sequential task planning. IEEE International Conference on
Robotics and Automation, 2023.
[1384] Zihan Zhou, Chong Li, Xinyi Chen, Shuo Wang, Yu Chao, Zhili Li, Haoyu Wang, Rongqiao An, Qi Shi,
Zhixing Tan, Xu Han, Xiaodong Shi, Zhiyuan Liu, and Maosong Sun. Llm×mapreduce: Simplified
long-sequence processing using large language models, arXiv preprint arXiv:2410.09342, 2024. URL
https://arxiv.org/abs/2410.09342v1.
[1385] Zijian Zhou, Ao Qu, Zhaoxuan Wu, Sunghwan Kim, Alok Prakash, Daniela Rus, Jinhua Zhao, B. Low,
and P. Liang. Mem1: Learning to synergize memory and reasoning for efficient long-horizon agents,
arXiv preprint arXiv:2506.15841, 2025. URL https://arxiv.org/abs/2506.15841v1.
[1386] Andrew Zhu, Liam Dugan, and Christopher Callison-Burch. Redel: A toolkit for llm-powered recursive
multi-agent systems. Conference on Empirical Methods in Natural Language Processing, 2024.
[1387] Dawei Zhu, Nan Yang, Liang Wang, Yifan Song, Wenhao Wu, Furu Wei, and Sujian Li. Pose: Efficient
context window extension of llms via positional skip-wise training. International Conference on
Learning Representations, 2023.
[1388] Hongyin Zhu. Metaaid 2.5: A secure framework for developing metaverse applications via large
language models. arXiv preprint, 2023.
[1389] Jason Zhu, Yanling Cui, Yuming Liu, Hao Sun, Xue Li, Markus Pelger, Liangjie Zhang, Tianqi Yan,
Ruofei Zhang, and Huasha Zhao. Textgnn: Improving text encoder via graph neural network in
sponsored search. The Web Conference, 2021.
[1390] Jiachen Zhu, Menghui Zhu, Renting Rui, Rong Shan, Congmin Zheng, Bo Chen, Yunjia Xi, Jianghao
Lin, Weiwen Liu, Ruiming Tang, Yong Yu, and Weinan Zhang. Evolutionary perspectives on the
evaluation of llm-based ai agents: A comprehensive survey, arXiv preprint arXiv:2506.11102, 2025.
URL https://arxiv.org/abs/2506.11102v1.
[1391] Jinguo Zhu, Weiyun Wang, Zhe Chen, Zhaoyang Liu, Shenglong Ye, Lixin Gu, Yuchen Duan, Hao
Tian, Weijie Su, Jie Shao, Zhangwei Gao, Erfei Cui, Yue Cao, Yangzhou Liu, Haomin Wang, Weiye
Xu, Hao Li, Jiahao Wang, Han Lv, Dengnian Chen, Songze Li, Yinan He, Tan Jiang, Jiapeng Luo,
Yi Wang, Cong He, Botian Shi, Xingcheng Zhang, Wenqi Shao, Junjun He, Ying Xiong, Wenwen Qu,
Peng Sun, Penglong Jiao, Lijun Wu, Kai Zhang, Hui Deng, Jiaye Ge, Kaiming Chen, Limin Wang, Min
Dou, Lewei Lu, Xizhou Zhu, Tong Lu, Dahua Lin, Yu Qiao, Jifeng Dai, and Wenhai Wang. Internvl3:
Exploring advanced training and test-time recipes for open-source multimodal models, arXiv preprint
arXiv:2504.10479v3, 2025. URL https://arxiv.org/abs/2504.10479v3.
[1392] Mingwei Zhu, Leigang Sha, Yu Shu, Kangjia Zhao, Tiancheng Zhao, and Jianwei Yin. Benchmarking
sequential visual input reasoning and prediction in multimodal large language models, arXiv preprint
arXiv:2310.13473v1, 2023. URL https://arxiv.org/abs/2310.13473v1.
[1393] Rongzhi Zhu, Xiangyu Liu, Zequn Sun, Yiwei Wang, and Wei Hu. Mitigating lost-in-retrieval problems
in retrieval augmented multi-hop question answering. 2025.
[1394] Rongzhi Zhu, Yi Liu, Zequn Sun, Yiwei Wang, and Wei Hu. When can large reasoning models save
thinking? mechanistic analysis of behavioral divergence in reasoning. 2025.
[1395] Runchuan Zhu, Zinco Jiang, Jiang Wu, Zhipeng Ma, Jiahe Song, Fengshuo Bai, Dahua Lin, Lijun Wu,
and Conghui He. Grait: Gradient-driven refusal-aware instruction tuning for effective hallucination
mitigation. 2025.
[1396] Runchuan Zhu, Zhipeng Ma, Jiang Wu, Junyuan Gao, Jiaqi Wang, Dahua Lin, and Conghui He.
Utilize the flow before stepping into the same river twice: Certainty represented knowledge flow
for refusal-aware instruction tuning. In Proceedings of the AAAI Conference on Artificial Intelligence,
volume 39, pages 26157–26165, 2025.
[1397] Tongyao Zhu, Qian Liu, L. Pang, Zhengbao Jiang, Min-Yen Kan, and Min Lin. Beyond memorization:
The challenge of random memory access in language models. Annual Meeting of the Association for
Computational Linguistics, 2024.
[1398] Xizhou Zhu, Yuntao Chen, Hao Tian, Chenxin Tao, Weijie Su, Chenyu Yang, Gao Huang, Bin Li,
Lewei Lu, Xiaogang Wang, Yu Qiao, Zhaoxiang Zhang, and Jifeng Dai. Ghost in the minecraft:
Generally capable agents for open-world environments via large language models with text-based
knowledge and memory, arXiv preprint arXiv:2305.17144, 2023. URL https://arxiv.org/
abs/2305.17144.
[1399] Yue Zhu, Hao Yu, Chen Wang, Zhuoran Liu, and Eun Kyung Lee. Towards efficient key-value cache
management for prefix prefilling in llm inference, arXiv preprint arXiv:2505.21919, 2025. URL
https://arxiv.org/abs/2505.21919v1.
[1400] Zhengqiu Zhu, Yong Zhao, Bin Chen, S. Qiu, Kai Xu, Quanjun Yin, Jin-Yu Huang, Zhong Liu, and Fei
Wang. Conversational crowdsensing: A parallel intelligence powered novel sensing approach, arXiv
preprint arXiv:2402.06654, 2024. URL https://arxiv.org/abs/2402.06654v1.
[1401] Zulun Zhu, Tiancheng Huang, Kai Wang, Junda Ye, Xinghe Chen, and Siqiang Luo. Graph-based
approaches and functionalities in retrieval-augmented generation: A comprehensive survey, arXiv
preprint arXiv:2504.10499, 2025. URL https://arxiv.org/abs/2504.10499v1.
[1402] Alex Zhuang, Ge Zhang, Tianyu Zheng, Xinrun Du, Junjie Wang, Weiming Ren, Stephen W. Huang,
Jie Fu, Xiang Yue, and Wenhu Chen. Structlm: Towards building generalist models for structured
knowledge grounding, arXiv preprint arXiv:2402.16671, 2024. URL https://arxiv.org/abs/
2402.16671v7.
[1403] Yuchen Zhuang, Yue Yu, Kuan Wang, Haotian Sun, and Chao Zhang. Toolqa: A dataset for llm
question answering with external tools. Neural Information Processing Systems, 2023.
[1404] Zhixiong Zhuang, Maria-Irina Nicolae, Hui-Po Wang, and Mario Fritz. Proxyprompt: Securing
system prompts against prompt extraction attacks, arXiv preprint arXiv:2505.11459, 2025. URL
https://arxiv.org/abs/2505.11459v1.
[1405] Ziyuan Zhuang, Zhiyang Zhang, Sitao Cheng, Fangkai Yang, Jia Liu, Shujian Huang, Qingwei Lin,
Saravan Rajmohan, Dongmei Zhang, and Qi Zhang. Efficientrag: Efficient retriever for multi-hop
question answering. In In Proceedings of the 2024 Conference on Empirical Methods in Natural
Language Processing, pages 3392–3411, 2024.
165
[1406] Chang Zong, Yuchen Yan, Weiming Lu, Eliot Huang, Jian Shao, and Y. Zhuang. Triad: A framework
leveraging a multi-role llm-based agent to solve knowledge base question answering. Conference on
Empirical Methods in Natural Language Processing, 2024.
[1407] Yongshuo Zong, Ondrej Bohdal, and Timothy M. Hospedales. Vl-icl bench: The devil in the details
of benchmarking multimodal in-context learning. arXiv preprint, 2024.
[1408] Yongshuo Zong, Ondrej Bohdal, and Timothy M. Hospedales. Vl-icl bench: The devil in the details
of multimodal in-context learning. International Conference on Learning Representations, 2024.
[1409] Henry Peng Zou, Wei-Chieh Huang, Yaozu Wu, Yankai Chen, Chunyu Miao, Hoang Nguyen, Yue
Zhou, Weizhi Zhang, Liancheng Fang, Langzhou He, Yangning Li, Yuwei Cao, Dongyuan Li, Renhe
Jiang, and Philip S. Yu. A survey on large language model based human-agent systems. arXiv
preprint, 2025.
[1410] Tao Zou, Le Yu, Yifei Huang, Leilei Sun, and Bo Du. Pretraining language models with text-attributed
heterogeneous graphs. Conference on Empirical Methods in Natural Language Processing, 2023.
[1411] Adam Zweiger, Jyothish Pari, Han Guo, Ekin Akyürek, Yoon Kim, and Pulkit Agrawal. Self-adapting
language models, arXiv preprint arXiv:2506.10943, 2025. URL https://arxiv.org/abs/2506.
10943v1.