语言大模型 / LLM - arXivDaily 专题

2310.10436 2024-05-27 cs.AI 88%

EconAgent: Large Language Model-Empowered Agents for Simulating Macroeconomic Activities

Nian Li, Chen Gao, Mingyu Li, Yong Li, Qingmin Liao

专题命中长上下文与记忆：large language model(title,abstract)；language model(title,abstract)；分类 cs.AI

Comments ACL 2024 (main conference)

2405.08848 2024-05-16 cs.SE cs.AI 88%

Automated Repair of AI Code with Large Language Models and Formal Verification

Yiannis Charalambous, Edoardo Manino, Lucas C. Cordeiro

专题命中长上下文与记忆：large language model(title,abstract)；language model(title,abstract)；分类 cs.AI

2310.08842 2024-05-08 cs.AI 88%

A Case-Based Persistent Memory for a Large Language Model

Ian Watson

专题命中长上下文与记忆：large language model(title,abstract)；language model(title,abstract)；分类 cs.AI

Comments 8 pages, 1 figure

2404.06345 2024-04-23 cs.AI cs.RO 88%

AgentsCoDriver: Large Language Model Empowered Collaborative Driving with Lifelong Learning

Senkang Hu, Zhengru Fang, Zihan Fang, Yiqin Deng, Xianhao Chen, Yuguang Fang

专题命中长上下文与记忆：large language model(title,abstract)；language model(title,abstract)；分类 cs.AI

2311.09182 2024-04-16 cs.CL 88%

ContraDoc: Understanding Self-Contradictions in Documents with Large Language Models

Jierui Li, Vipul Raheja, Dhruv Kumar

专题命中长上下文与记忆：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL

Comments Accepted to NAACL 2024 main conference

2310.16450 2024-03-26 cs.CL 88%

CLEX: Continuous Length Extrapolation for Large Language Models

Guanzheng Chen, Xin Li, Zaiqiao Meng, Shangsong Liang, Lidong Bing

专题命中长上下文与记忆：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL

Comments ICLR 2024

2312.09571 2023-12-18 cs.CL cs.IT math.IT 88%

Extending Context Window of Large Language Models via Semantic Compression

Weizhi Fei, Xueyan Niu, Pingyi Zhou, Lu Hou, Bo Bai, Lei Deng, Wei Han

专题命中长上下文与记忆：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL

2310.14248 2023-10-24 cs.CL 88%

From Static to Dynamic: A Continual Learning Framework for Large Language Models

Mingzhe Du, Anh Tuan Luu, Bin Ji, See-kiong Ng

专题命中长上下文与记忆：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL

2310.11634 2023-10-19 cs.CL 88%

MAGNIFICo: Evaluating the In-Context Learning Ability of Large Language Models to Generalize to Novel Interpretations

Arkil Patel, Satwik Bhattamishra, Siva Reddy, Dzmitry Bahdanau

专题命中长上下文与记忆：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL

Comments EMNLP 2023

2304.05511 2023-04-13 cs.LG 88%

Training Large Language Models Efficiently with Sparsity and Dataflow

Venkat Srinivasan, Darshan Gandhi, Urmish Thakker, Raghu Prabhakar

专题命中长上下文与记忆：large language model(title,abstract)；language model(title,abstract)；分类 cs.LG

2301.04589 2023-01-12 cs.CL cs.FL 88%

Memory Augmented Large Language Models are Computationally Universal

Dale Schuurmans

专题命中长上下文与记忆：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL

Comments 23 pages, 0 figures

2205.11503 2022-05-24 cs.CL 88%

Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models

Mirac Suzgun, Luke Melas-Kyriazi, Dan Jurafsky

专题命中长上下文与记忆：language model(title,abstract)；small language model(title)；prompting(abstract)；分类 cs.CL

Comments GitHub page: https://github.com/suzgunmirac/prompt-and-rerank. Project page: https://lukemelas.github.io/prompt-and-rerank/

2608.16168 2026-08-18 cs.CL cs.AI 新提交 88%

QUMem: Personalized Memory for Query-Conditioned User-State Inference in LLM Agents

QUMem：面向大语言模型智能体查询条件用户状态推理的个性化记忆

Heng Wang, Yifei Li, Lingling Zhang, Pengyu Li, Xinyu Che, Xinyu Zhang, Zesheng Yang

专题命中长上下文与记忆：LLM(title,abstract)；large language model(abstract)；language model(abstract)；分类 cs.CL、cs.AI

AI总结 QUMem 是一种结构化记忆框架，通过分割交互历史、分解记忆类型、多查询检索等方法，在 PersonaMem 和 KnowU-Bench 上实现了查询条件用户状态推理的最优性能，提升了大语言模型智能体的长期个性化能力。

Comments 9pages,3figures

2508.04412 2026-08-11 cs.AI cs.CL cs.HC 版本更新 88%

Beyond Pixels: Exploring DOM Downsampling for LLM-Based Web Agents

超越像素：探索面向基于大语言模型的智能体的DOM下采样

Thassilo M. Schiepanski, Nicholas Piël

专题命中长上下文与记忆：LLM(title,abstract)；large language model(abstract)；language model(abstract)；分类 cs.CL、cs.AI

AI总结本文提出D2Snap算法对DOM下采样，解决原始DOM令牌占用过大的问题，经评估其在GPT-4o智能体上的成功率达73%，接近基于GUI快照的基线性能。

Comments 21 pages, LaTeX, print version; enhanced algorithm, settled on non-inferiority claim, added downsampling monotonicity and linearity results

2608.07438 2026-08-10 cs.AI cs.CL cs.HC 新提交 88%

PsychoAgent: An Affect-Sensitive Cognitive Architecture for Conflict-Aware Memory in LLM Agents

PsychoAgent：面向大语言模型智能体的、具备冲突感知记忆的情感敏感认知架构

Mohammad Amanlou, Parham Abed Azad, Farbod Davoodi, Mostafa Masumi, Behnam Bahrak, Abdol-Hossein Vahabie

机构 * School of Electrical and Computer Engineering, College of Engineering, University of Tehran（德黑兰大学工程学院电气与计算机工程学院）； Sharif University of Technology（谢里夫理工大学）； Missouri University of Science and Technology（密苏里科技大学）； Tehran Institute for Advanced Studies, Khatam University（哈塔姆大学德黑兰高等研究院）

专题命中长上下文与记忆：LLM(title,summary_cn)；分类 cs.CL、cs.AI

AI总结该研究提出PsychoAgent架构，分离事实与情感记忆并通过冲突感知控制器整合，在三类冲突场景中其冲突关键记忆检索率优于相关基线，为LLM智能体建模类人冲突效应提供了可检查机制。

Comments 12 pages main paper + 10 pages supplementary material; supplementary material included

2608.03130 2026-08-05 cs.CR cs.CL cs.LG 新提交 88%

DP-MemView: A Memory Interface for Attribute-Level Transcript Privacy in Long-Term LLM Agents

DP-MemView：面向长期大语言模型智能体的属性级对话记录隐私的内存接口

Jong Wook Kim, Byoungjae Min, Kennedy Edemacu, Yoonhyuk Choi, Sae-Hong Cho, Beakcheol Jang

专题命中长上下文与记忆：LLM(title,summary_cn)；分类 cs.CL、cs.LG

AI总结针对长期LLM智能体的自适应对话记录隐私威胁，提出差分隐私内存接口DP-MemView，经实验验证可在保留个性化和响应质量的同时降低隐私泄露风险。

Comments 18 pages, 2 figures, 9 tables

2605.08442 2026-08-05 cs.CR cs.AI cs.LG 版本更新 88%

Injection-Execution Dissociation: A Mechanistic Evaluation of Persistent Memory Attacks and Defenses in Stateful LLM Agents

多层架构中的防御效果：对持久内存攻击状态机LLM代理的机制性评估

Jun Wen Leong

专题命中长上下文与记忆：LLM(title,title_cn)；分类 cs.AI、cs.LG

AI总结本文评估了六种防御措施在九个开源模型上的延迟触发攻击效果，发现输入级和检索级过滤器效果不佳，而内存层工具门控（Memory Sandbox）显著降低攻击成功率，揭示了不同防御类别的失效原因。

Comments v4: Added double dissociation (reasoning-mode ablation), content-layer defense (RATG), loaded-corpus frontier evaluation (21 models, 3 providers, N=40), 7B judge capability bound, reproducibility validity criterion, ethics/disclosure statement. Gemini 3.1 Pro Preview 95% ASR; GPT-5.1 regression (22.5%); tripartite vendor divergence. Code: github.com/junwenleong/stateful-agent-security-eval

2607.23286 2026-07-28 cs.AI cs.LG 新提交 88%

TopoFE: topology-aware LLM-guided Automated Feature Engineering

TopoFE：拓扑感知的大语言模型引导的自动特征工程

Sha Li, Naren Ramakrishnan

机构 * Virginia Tech（弗吉尼亚理工大学）

专题命中长上下文与记忆：LLM(title,abstract)；large language model(abstract)；language model(abstract)；分类 cs.AI、cs.LG

AI总结研究表格学习中自动特征工程问题，提出拓扑感知多岛进化框架TOPOFE，结合特定族探索、自适应提示记忆和拓扑引导知识转移，在多个公共数据集实验中，相比现有方法有改进，能发现更多样可转移特征程序。

2601.00679 2026-07-07 cs.NE cs.AI cs.LG 88%

QSLM: A Performance- and Memory-aware Quantization Framework with Tiered Search Strategy for Spike-driven Language Models

QSLM：一种性能和内存感知的量化框架，具有分层搜索策略用于脉冲驱动语言模型

Rachmad Vidya Wicaksana Putra, Pasindu Wickramasinghe, Muhammad Shafique

机构 * eBRAIN Lab, New York University (NYU) Abu Dhabi（eBRAIN实验室，纽约大学阿布扎比分校）

专题命中长上下文与记忆：language model(title,abstract)；SLM(abstract,abstract_cn)；large language model(abstract)；分类 cs.AI、cs.LG

AI总结本文提出QSLM框架，通过分层搜索策略和多目标性能-内存权衡函数，实现对预训练脉冲驱动语言模型的自动量化压缩，减少内存占用并降低能耗，同时保持高性能。

Comments Accepted at the Design, Automation and Test in Europe Conference (DATE) 2025 on April 20th-22nd, 2026 in Verona, Italy

2607.02255 2026-07-03 cs.AI cs.CL 新提交 88%

AgenticSTS: A Bounded-Memory Testbed for Long-Horizon LLM Agents

AgenticSTS: 一种用于长周期LLM智能体的有界记忆测试平台

Xiangchen Cheng, Yunwei Jiang, Jianwen Sun, Zizhen Li, Chuanhao Li, Xiangcheng Cao, Yihao Liu, Fanrui Zhang, Li Jin, Kaipeng Zhang

机构 * Alaya Lab（Alaya实验室）； Shanghai Jiao Tong University（上海交通大学）； Shanghai Innovation Institute（上海创新研究院）； Nankai University（南开大学）； University of Science and Technology of China（中国科学技术大学）

专题命中长上下文与记忆：LLM(title,title_cn)；分类 cs.CL、cs.AI

AI总结提出一种有界记忆契约，通过类型化检索组装用户消息，避免原始跨决策转录，在《杀戮尖塔2》中实现并验证，揭示了战略技能层对胜率的提升效果。

2606.07552 2026-06-26 cs.MA cs.AI cs.LG 新提交 88%

Symbolic Reasoning Frameworks Trigger Memory-Mediated Ecosystem Dynamics in Multi-Agent LLM Systems

符号推理框架在多智能体战略环境中调节大语言模型的风险规避

Augustin Chan

机构 * iterative.day

专题命中长上下文与记忆：LLM(title,summary_cn)；large language model(abstract)；language model(abstract)；分类 cs.AI、cs.LG

AI总结本研究通过注入符号推理框架（如易经、塔罗牌）作为反思提示，发现其能差异化调节LLM的风险规避倾向，并在多智能体博弈中产生框架特定的胜者分布，且该效应源于反思过程而非内容遵循。

Comments 28 pages, 4 figures, 9 tables, 6 listings. Code and data: https://doi.org/10.5281/zenodo.20338937

2601.03785 2026-06-25 cs.CL cs.AI 版本更新 88%

Membox: Weaving Topic Continuity into Long-Range Memory for LLM Agents

Membox: 将主题连续性编织到LLM智能体的长期记忆中

Dehao Tao, Guoliang Ma, Yongfeng Huang, Minghu Jiang

机构 * Tsinghua University（清华大学）； Xinjiang University（新疆大学）

专题命中长上下文与记忆：LLM(title,title_cn)；分类 cs.CL、cs.AI

AI总结提出Membox层次记忆架构，通过主题织机按局部主题组织对话，并通过轨迹织机跨会话链接宏主题轨迹，在LoCoMo和DialSim上显著提升长程记忆检索性能。

2606.10677 2026-06-10 cs.AI cs.CL 新提交 88%

Infini Memory: Maintainable Topic Documents for Long-Term LLM Agent Memory

Infini Memory：用于长期LLM智能体记忆的可维护主题文档

Suozhao Ji, Baodong Wu, Zehao Wang, Lei Xia, Qingping Li, Ruisong Wang, Wenbo Ding, Zhenhua Zhu, Boxun Li, Guohao Dai, Yu Wang

机构 * Infinigence AI（InfiniGen AI）； Tsinghua University（清华大学）； Shanghai Jiaotong University（上海交通大学）

专题命中长上下文与记忆：LLM(title,title_cn)；分类 cs.CL、cs.AI

AI总结提出Infini Memory架构，将智能体记忆组织为主题文档，通过缓冲合并和迭代检索实现可维护的长期记忆，在MemoryAgentBench上达到64.7%的总体得分。

2606.09483 2026-06-09 cs.CL cs.AI 新提交 88%

Memory Beyond Recall: A Dual-Process Cognitive Memory System for Self-Evolving LLM Agents

超越回忆的记忆：用于自进化LLM代理的双过程认知记忆系统

Tianxiang Fei, Mingyang Song, Mao Zheng, Xiang Yu

机构 * Tencent（腾讯）

专题命中长上下文与记忆：LLM(title,title_cn)；分类 cs.CL、cs.AI

AI总结提出DCPM系统，基于双过程理论将代理记忆组织为认知能力层次，通过同步日间写入器和异步夜间引擎分别处理信念修正和模式归纳，在隐式跨会话推理任务上提升显著。

2505.11166 2026-06-04 cs.CL cs.AI 88%

SoLoPO: Unlocking Long-Context Capabilities in LLMs via Short-to-Long Preference Optimization

SoLoPO: 通过短到长偏好优化解锁大语言模型的长上下文能力

Huashan Sun, Shengyi Liao, Yansen Han, Yu Bai, Yang Gao, Cheng Fu, Weizhou Shen, Fanqi Wan, Ming Yan, Ji Zhang, Fei Huang

机构 * Tongyi Lab, Alibaba Group（通义实验室，阿里巴巴集团）

专题命中长上下文与记忆：preference optimization(title,abstract)；large language model(abstract)；language model(abstract)；pretraining(abstract)

AI总结提出SoLoPO框架，将长上下文偏好优化解耦为短上下文偏好优化和短到长奖励对齐，以提升大语言模型的长上下文利用能力。

Comments Published as a conference paper at ICLR 2026

2605.27489 2026-05-28 cs.CR cs.AI cs.LG 88%

HARP: Measuring Harm Amplification in Multi-Agent LLM Systems

HARP: 多智能体大语言模型系统中的危害放大测量

Md Hafizur Rahman, Zafaryab Haider, Tanzim Mahfuz, Prabuddha Chakraborty

机构 * Electrical and Computer Engineering University of Maine（电气与计算机工程大学缅因州大学）

专题命中长上下文与记忆：LLM(title,summary_cn)；分类 cs.AI、cs.LG

AI总结提出HARP方法，通过比较清洁与扰动执行轨迹，量化多智能体LLM系统中局部扰动如何传播为全局危害，并在金融七智能体系统中验证了不同攻击和防御的效果。

Comments 39 pages, 12 figures, 12 tables, and 1 algorithm

2605.24786 2026-05-26 cs.LG cs.AI 88%

CONF-KV: Confidence-Aware KV Cache Eviction with Mixed-Precision Storage for Long-Horizon LLM

CONF-KV：面向长序列LLM的置信度感知KV缓存淘汰与混合精度存储

Yubo Li, Yidi Miao

机构 * Carnegie Mellon University（卡内基梅隆大学）

专题命中长上下文与记忆：LLM(title,title_cn)；分类 cs.AI、cs.LG

AI总结提出CONF-KV方法，利用模型当前不确定性（置信度）动态调整KV缓存预算，结合混合精度存储和分块在线softmax注意力，在长序列推理中显著降低显存占用并保持高精度。

2604.12376 2026-05-26 cs.CL cs.AI 88%

Cooperative Memory Paging with Keyword Bookmarks for Long-Horizon LLM Conversations

面向长程LLM对话的协作式内存分页与关键词书签

Ziyang Liu

专题命中长上下文与记忆：LLM(title,title_cn)；分类 cs.CL、cs.AI

AI总结提出协作式分页方法，用关键词书签替代被驱逐的对话片段，并赋予模型 recall() 工具按需检索，在 LoCoMo 基准上四个模型均取得最佳答案质量，并通过消融实验揭示分页设计的关键因素。

Comments The authors have decided to withdraw this version following internal review regarding authorship and contribution agreements

2512.24601 2026-05-12 cs.AI cs.CL 88%

Recursive Language Models

递归语言模型

Alex L. Zhang, Tim Kraska, Omar Khattab

机构 * MIT CSAIL（麻省理工学院计算机科学与人工智能实验室）

专题命中长上下文与记忆：language model(title,abstract)；LLM(abstract,abstract_cn)；large language model(abstract)；分类 cs.CL、cs.AI

AI总结本文提出递归语言模型（RLMs），通过将长提示视为外部环境的一部分，使大语言模型能够递归调用自身处理长输入，从而在多个长上下文任务中显著提升性能。

Comments 9 pages, 43 with Appendix

2605.07363 2026-05-11 cs.LG cs.AI 88%

MISA: Mixture of Indexer Sparse Attention for Long-Context LLM Inference

MISA：混合索引器稀疏注意力用于长上下文LLM推理

Ruijie Zhou, Fanxu Meng, Yufei Xu, Tongxuan Liu, Guangming Lu, Muhan Zhang, Wenjie Pei

机构 * PKU（北京大学）

专题命中长上下文与记忆：LLM(title,title_cn)；分类 cs.AI、cs.LG

AI总结 MISA通过将DSA索引器头部视为专家混合池，降低长上下文LLM推理的计算成本，同时保持性能，比HISA更优。

Comments https://github.com/MuLabPKU/TransArch

1. 长上下文与记忆 4797 篇

EconAgent: Large Language Model-Empowered Agents for Simulating Macroeconomic Activities

Automated Repair of AI Code with Large Language Models and Formal Verification

A Case-Based Persistent Memory for a Large Language Model

AgentsCoDriver: Large Language Model Empowered Collaborative Driving with Lifelong Learning

ContraDoc: Understanding Self-Contradictions in Documents with Large Language Models

CLEX: Continuous Length Extrapolation for Large Language Models

Extending Context Window of Large Language Models via Semantic Compression

From Static to Dynamic: A Continual Learning Framework for Large Language Models

MAGNIFICo: Evaluating the In-Context Learning Ability of Large Language Models to Generalize to Novel Interpretations

Training Large Language Models Efficiently with Sparsity and Dataflow

Memory Augmented Large Language Models are Computationally Universal

Prompt-and-Rerank: A Method for Zero-Shot and Few-Shot Arbitrary Textual Style Transfer with Small Language Models

QUMem: Personalized Memory for Query-Conditioned User-State Inference in LLM Agents

Beyond Pixels: Exploring DOM Downsampling for LLM-Based Web Agents

PsychoAgent: An Affect-Sensitive Cognitive Architecture for Conflict-Aware Memory in LLM Agents

DP-MemView: A Memory Interface for Attribute-Level Transcript Privacy in Long-Term LLM Agents

Injection-Execution Dissociation: A Mechanistic Evaluation of Persistent Memory Attacks and Defenses in Stateful LLM Agents

TopoFE: topology-aware LLM-guided Automated Feature Engineering

QSLM: A Performance- and Memory-aware Quantization Framework with Tiered Search Strategy for Spike-driven Language Models

AgenticSTS: A Bounded-Memory Testbed for Long-Horizon LLM Agents

Symbolic Reasoning Frameworks Trigger Memory-Mediated Ecosystem Dynamics in Multi-Agent LLM Systems

Membox: Weaving Topic Continuity into Long-Range Memory for LLM Agents

Infini Memory: Maintainable Topic Documents for Long-Term LLM Agent Memory

Memory Beyond Recall: A Dual-Process Cognitive Memory System for Self-Evolving LLM Agents

SoLoPO: Unlocking Long-Context Capabilities in LLMs via Short-to-Long Preference Optimization

HARP: Measuring Harm Amplification in Multi-Agent LLM Systems

CONF-KV: Confidence-Aware KV Cache Eviction with Mixed-Precision Storage for Long-Horizon LLM

Cooperative Memory Paging with Keyword Bookmarks for Long-Horizon LLM Conversations

Recursive Language Models

MISA: Mixture of Indexer Sparse Attention for Long-Context LLM Inference