语言大模型 / LLM - arXivDaily 专题

2603.24257 2026-03-31 cs.CV 82%

Memory-Augmented Vision-Language Agents for Persistent and Semantically Consistent Object Captioning

具有记忆增强的视觉-语言代理用于持久且语义一致的对象描述

Tommaso Galliena, Stefano Rosa, Tommaso Apicella, Pietro Morerio, Alessio Del Bue, Lorenzo Natale

机构 * Italian Institute of Technology, Genoa, Italy（意大利理工学院，热那亚，意大利）； University of Genoa, Genoa, Italy（热那亚大学，热那亚，意大利）

专题命中长上下文与记忆：language agent(title,abstract)；language model(abstract)

AI总结本文提出一种统一的记忆增强视觉-语言代理，通过单一自回归框架同时处理数据关联、对象描述和探索策略，提升持久且语义一致的对象描述能力。

Comments 24 pages, 7 figures, 7 tables (including Supplementary Materials)

2603.24837 2026-03-27 cs.CR 82%

Bridging Code Property Graphs and Language Models for Program Analysis

连接代码属性图与语言模型用于程序分析

Ahmed Lekssays

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

AI总结本文提出codebadger，结合Joern的代码属性图引擎与LLM，解决代码分析中的挑战，通过高级工具实现程序切片、污点追踪等，提升大规模代码库的分析能力。

Comments Accepted at Software Vulnerability Management Workshop @ ICSE 2026

2603.21576 2026-03-26 physics.optics cs.AI cs.AR cs.CL cs.LG 82%

PRISM: Breaking the O(n) Memory Wall in Long-Context LLM Inference via O(1) Photonic Block Selection

PRISM：通过O(1)光子块选择突破长上下文LLM推理的O(n)内存瓶颈

Hyoseok Park, Yeonsang Park

机构 * Department of Physics, Chungnam National University（Chungnam National University 物理系）

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

AI总结 PRISM通过光子广播与加权范式，实现长上下文LLM推理中O(n)内存带宽的突破，采用O(1)光子块选择方法，显著提升效率与能效比。

Comments 28 pages, 27 figures, 15 tables, including supplementary material. Code available at https://github.com/hyoseokp/PRISM

2603.15653 2026-03-18 cs.CL cs.AI cs.LG 82%

Recursive Language Models Meet Uncertainty: The Surprising Effectiveness of Self-Reflective Program Search for Long Context

递归语言模型与不确定性：自我反思程序搜索在长上下文中的意外效果

Keivan Alizadeh, Parshin Shojaee, Minsik Cho, Mehrdad Farajtabar

机构 * Apple（苹果公司）

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

AI总结本文提出SRLM框架，通过引入不确定性感知的自我反思，提升长上下文处理能力，实验显示其在多种任务中优于现有基线，且无需递归机制。

Comments preprint

2512.17023 2026-03-16 cs.DC cs.SE 82%

LLM-HPC++: Evaluating LLM-Generated Modern C++ and MPI+OpenMP Codes for Scalable Mandelbrot Set Computation

LLM-HPC++: 评估LLM生成的现代C++和MPI+OpenMP代码用于可扩展的曼德勃罗集计算

Patrick Diehl, Noujoud Nader, Deepti Gupta

专题命中长上下文与记忆：LLM(title)；large language model(abstract)；language model(abstract)

AI总结本文评估了LLM生成的C++和MPI+OpenMP代码在曼德勃罗集计算中的性能，发现ChatGPT-4和5在语法精度和可扩展性方面表现突出。

2603.03290 2026-03-05 cs.CL cs.AI cs.IR cs.LG 82%

AriadneMem: Threading the Maze of Lifelong Memory for LLM Agents

AriadneMem: 为LLM代理梳理终身记忆的迷宫

Wenhui Zhu, Xiwen Chen, Zhipeng Wang, Jingjing Wang, Xuanzhao Dong, Minzhou Huang, Rui Cai, Hejian Sang, Hao Wang, Peijie Qiu, Yueyue Deng, Prayag Tiwari, Brendan Hogan Rappazzo, Yalin Wang

机构 * Arizona State University（亚利桑那州立大学）； Morgan Stanley（摩根大通）； Rice University（里奇大学）； Clemson University（克莱姆森大学）； Northwestern University（西北大学）； UC Davis（加州大学戴维斯分校）； Iowa State University（爱荷华州立大学）； Washington University in St. Louis（圣路易斯华盛顿大学）； Columbia University（哥伦比亚大学）； Halmstad University（哈马格大学）

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

AI总结 AriadneMem通过解耦的两阶段流程提升LLM代理在长期对话中的多跳和平均F1表现，同时减少运行时间。

2510.12064 2026-02-27 cs.NI 82%

GeoPipe: a Geo-distributed LLM Training Framework with enhanced Pipeline Parallelism in a Lossless RDMA-enabled Datacenter Optical Transport Network

GeoPipe：一种具有增强流水线并行性的地理分布式LLM训练框架，在无损RDMA启用的数据中心光传送网络中

Jun Dai, Xiaorun Wang, Kexiong Fang, Zheng Yang, Yuefeng Ji, Jiawei Zhang

专题命中长上下文与记忆：LLM(title)；large language model(abstract)；language model(abstract)

AI总结 GeoPipe提出了一种基于无损RDMA的数据中心光传送网络的地理分布式LLM训练框架，通过增强的流水线并行技术显著降低计算气泡比率。

Comments 6 pages, 4 figures

Journal ref 2025 Asia Communications and Photonics Conference (ACP), 2025

2602.00746 2026-02-03 cs.SE cs.CV 82%

Can Vision-Language Models Handle Long-Context Code? An Empirical Study on Visual Compression

视觉-语言模型能处理长上下文代码吗？对视觉压缩的实证研究

Jianping Zhong, Guochang Li, Chen Zhi, Junxiao Han, Zhen Qin, Xinkui Zhao, Nan Wang, Shuiguang Deng, Jianwei Yin

机构 * Zhejiang University（浙江大学）； Hangzhou City University（杭州市大学）； Shenzhou Aerospace Software Technology Company Limited（神州航天软件技术有限公司）

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

AI总结本文提出LongCodeOCR，通过视觉压缩替代文本压缩，提升长上下文代码任务的性能与效率，同时揭示了视觉压缩在全局依赖支持与保真度之间的权衡。

2601.13631 2026-01-21 cs.OS cs.DC 82%

ContiguousKV: Accelerating LLM Prefill with Granularity-Aligned KV Cache Management

ContiguousKV: 通过粒度对齐的KV缓存管理加速LLM预填

Jing Zou, Shangyu Wu, Hancong Duan, Qiao Li, Chun Jason Xue

专题命中长上下文与记忆：LLM(title)；large language model(abstract)；language model(abstract)

AI总结 ContiguousKV通过粒度对齐的KV缓存管理，提升LLM预填效率，实现3.85倍加速并保持输出质量。

2511.21037 2025-11-27 cs.HC 82%

LOOM: Personalized Learning Informed by Daily LLM Conversations Toward Long-Term Mastery via a Dynamic Learner Memory Graph

LOOM：通过动态学习者记忆图实现每日LLM对话指导的个性化学习以实现长期掌握

Justin Cui, Kevin Pu, Tovi Grossman

专题命中长上下文与记忆：LLM(title,abstract)；foundation model(abstract)

AI总结 LOOM通过动态学习者记忆图分析每日LLM对话，生成个性化学习材料以实现长期掌握，兼顾连续性和主动性，提升学习者持续进步与即时相关性的平衡。

Comments Accepted to the PerFM Workshop at AAAI 2026

2507.03305 2025-10-31 cs.DC 82%

Analysis and Optimized CXL-Attached Memory Allocation for Long-Context LLM Fine-Tuning

Yong-Cheng Liaw, Shuo-Han Chen

专题命中长上下文与记忆：LLM(title)；large language model(abstract)；language model(abstract)

Comments 13 pages, 15 figures, 2 tables

2510.24726 2025-10-30 eess.SP cs.CY stat.AP 82%

Modelling Real-Life Cycling Decisions in Real Urban Settings Through Psychophysiology and LLM-Derived Contextual Data

Maximiliano Rosadio Z., Angel Jimenez-Molina, Bastián Henríquez, Paulina Leiva, Ricardo Hurtubia, Ricardo De La Paz Guala, Leandro Gayozo, C. Angelo Guevara

专题命中长上下文与记忆：LLM(title)；large language model(abstract)；language model(abstract)

Comments 31 pages, 10 figures

2505.15962 2025-10-06 cs.CL cs.AI cs.LG 82%

Pre-training Limited Memory Language Models with Internal and External Knowledge

Linxi Zhao, Sofian Zalouk, Christian K. Belardi, Justin Lovelace, Jin Peng Zhou, Ryan Thomas Noonan, Dongyoung Go, Kilian Q. Weinberger, Yoav Artzi, Jennifer J. Sun

机构 * Department of Computer Science Cornell University（计算机科学系哥伦比亚大学）

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

Comments Code, models, and data available at https://github.com/kilian-group/LMLM

2509.14223 2025-09-23 cs.LG cs.AI cs.CL 82%

Fresh in memory: Training-order recency is linearly encoded in language model activations

Dmitrii Krasheninnikov, Richard E. Turner, David Krueger

机构 * University of Cambridge（剑桥大学）； Mila, University of Montreal（蒙特利尔大学米尔研究所）

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

2508.16629 2025-08-26 cs.LG cs.AI cs.CL cs.IR 82%

Learn to Memorize: Optimizing LLM-based Agents with Adaptive Memory Framework

Zeyu Zhang, Quanyu Dai, Rui Li, Xiaohe Bo, Xu Chen, Zhenhua Dong

机构 * Gaoling School of Artificial Intelligence, Renmin University of China（中国人民大学人工智能学院）； Huawei Noah’s Ark Lab（华为诺亚实验室）

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

Comments 17 pages, 4 figures, 5 tables

2508.07026 2025-08-12 quant-ph 82%

Bridging Classical and Quantum Computing for Next-Generation Language Models

Yi Pan, Hanqi Jiang, Junhao Chen, Yiwei Li, Huaqin Zhao, Lin Zhao, Yohannes Abate, Yingfeng Wang, Tianming Liu

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

2508.02679 2025-08-11 cs.HC 82%

LLM Agent-Based Simulation of Student Activities and Mental Health Using Smartphone Sensing Data

Wayupuk Sommuang, Kun Kerdthaisong, Pasin Buakhaw, Aslan B. Wong, Nutchanon Yongsatianchot

专题命中长上下文与记忆：LLM(title,abstract)；prompting(abstract)

2503.02247 2025-07-22 cs.CV cs.RO 82%

WMNav: Integrating Vision-Language Models into World Models for Object Goal Navigation

Dujun Nie, Xianda Guo, Yiqun Duan, Ruijun Zhang, Long Chen

机构 * Institute of Automation, Chinese Academy of Sciences（中国科学院自动化研究所）； School of Computer Science, Wuhan University（武汉大学计算机学院）； School of Computer Science, University of Technology Sydney（悉尼大学计算机学院）； IAIR, Xi’an Jiaotong University（西安交通大学IAIR）

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

Comments 8 pages, 5 figures

Journal ref IROS 2025

2412.16926 2025-05-29 cs.CL cs.AI cs.LG 82%

Revisiting In-Context Learning with Long Context Language Models

Jinheon Baek, Sun Jae Lee, Prakhar Gupta, Geunseob Oh, Siddharth Dalmia, Prateek Kolhar

机构 * KAIST（韩国科学技术院）； Google DeepMind（谷歌DeepMind）

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

Comments ACL Findings 2025

2505.18279 2025-05-27 cs.MA cs.AI cs.CL cs.LG 82%

Collaborative Memory: Multi-User Memory Sharing in LLM Agents with Dynamic Access Control

Alireza Rezazadeh, Zichao Li, Ange Lou, Yuying Zhao, Wei Wei, Yujia Bao

机构 * Center for Advanced AI, Accenture（Accenture高级人工智能研究中心）

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

2505.11888 2025-05-20 cs.HC 82%

AR Secretary Agent: Real-time Memory Augmentation via LLM-powered Augmented Reality Glasses

Raphaël A. El Haddad, Zeyu Wang, Yeonsu Shin, Ranyi Liu, Yuntao Wang, Chun Yu

专题命中长上下文与记忆：LLM(title)；large language model(abstract)；language model(abstract)

2412.17593 2025-04-30 cs.IR 82%

Leveraging Memory Retrieval to Enhance LLM-based Generative Recommendation

Chengbing Wang, Yang Zhang, Fengbin Zhu, Jizhi Zhang, Tianhao Shi, Fuli Feng

专题命中长上下文与记忆：LLM(title)；large language model(abstract)；language model(abstract)

Comments Accepted by WWW'2025

2504.12526 2025-04-18 cs.LG cs.AI cs.CL 82%

MOM: Memory-Efficient Offloaded Mini-Sequence Inference for Long Context Language Models

Junyang Zhang, Tianyi Zhu, Cheng Luo, Anima Anandkumar

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

Comments Submitted to COLM

2410.10781 2025-03-04 cs.CL cs.AI cs.LG 82%

When Attention Sink Emerges in Language Models: An Empirical View

Xiangming Gu, Tianyu Pang, Chao Du, Qian Liu, Fengzhuo Zhang, Cunxiao Du, Ye Wang, Min Lin

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

Comments ICLR 2025 (Spotlight)

2502.18465 2025-02-27 cs.SE 82%

Empirical Research on Utilizing LLM-based Agents for Automated Bug Fixing via LangGraph

Jialin Wang, Zhihua Duan

专题命中长上下文与记忆：LLM(title)；large language model(abstract)；language model(abstract)

2410.04739 2024-12-30 cs.CL cs.AI cs.IR cs.LG 82%

TableRAG: Million-Token Table Understanding with Language Models

Si-An Chen, Lesly Miculicich, Julian Martin Eisenschlos, Zifeng Wang, Zilong Wang, Yanfei Chen, Yasuhisa Fujii, Hsuan-Tien Lin, Chen-Yu Lee, Tomas Pfister

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

Comments Accepted to NeurIPS 2024

2407.16695 2024-12-04 cs.CL cs.AI cs.LG 82%

Stress-Testing Long-Context Language Models with Lifelong ICL and Task Haystack

Xiaoyue Xu, Qinyuan Ye, Xiang Ren

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

Comments NeurIPS 2024 (Datasets and Benchmarks Track). Code: https://github.com/INK-USC/Lifelong-ICL Website: https://inklab.usc.edu/lifelong-icl/

2406.19508 2024-07-24 cs.SE 82%

Code Linting using Language Models

Darren Holden, Nafiseh Kahani

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

Comments 44 pages, 3 figures

2407.00020 2024-07-02 cs.CV cs.AI cs.CL cs.IT cs.LG math.IT 82%

Visual Language Model based Cross-modal Semantic Communication Systems

Feibo Jiang, Chuanguo Tang, Li Dong, Kezhi Wang, Kun Yang, Cunhua Pan

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

Comments 12 pages, 10 figures

2406.02080 2024-06-05 cs.CL cs.AI cs.LG math.DS 82%

LongSSM: On the Length Extension of State-space Models in Language Modelling

Shida Wang

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

Comments 23 pages

1. 长上下文与记忆 4771 篇

Memory-Augmented Vision-Language Agents for Persistent and Semantically Consistent Object Captioning

Bridging Code Property Graphs and Language Models for Program Analysis

PRISM: Breaking the O(n) Memory Wall in Long-Context LLM Inference via O(1) Photonic Block Selection

Recursive Language Models Meet Uncertainty: The Surprising Effectiveness of Self-Reflective Program Search for Long Context

LLM-HPC++: Evaluating LLM-Generated Modern C++ and MPI+OpenMP Codes for Scalable Mandelbrot Set Computation

AriadneMem: Threading the Maze of Lifelong Memory for LLM Agents

GeoPipe: a Geo-distributed LLM Training Framework with enhanced Pipeline Parallelism in a Lossless RDMA-enabled Datacenter Optical Transport Network

Can Vision-Language Models Handle Long-Context Code? An Empirical Study on Visual Compression

ContiguousKV: Accelerating LLM Prefill with Granularity-Aligned KV Cache Management

LOOM: Personalized Learning Informed by Daily LLM Conversations Toward Long-Term Mastery via a Dynamic Learner Memory Graph

Analysis and Optimized CXL-Attached Memory Allocation for Long-Context LLM Fine-Tuning

Modelling Real-Life Cycling Decisions in Real Urban Settings Through Psychophysiology and LLM-Derived Contextual Data

Pre-training Limited Memory Language Models with Internal and External Knowledge

Fresh in memory: Training-order recency is linearly encoded in language model activations

Learn to Memorize: Optimizing LLM-based Agents with Adaptive Memory Framework

Bridging Classical and Quantum Computing for Next-Generation Language Models

LLM Agent-Based Simulation of Student Activities and Mental Health Using Smartphone Sensing Data

WMNav: Integrating Vision-Language Models into World Models for Object Goal Navigation

Revisiting In-Context Learning with Long Context Language Models

Collaborative Memory: Multi-User Memory Sharing in LLM Agents with Dynamic Access Control

AR Secretary Agent: Real-time Memory Augmentation via LLM-powered Augmented Reality Glasses

Leveraging Memory Retrieval to Enhance LLM-based Generative Recommendation

MOM: Memory-Efficient Offloaded Mini-Sequence Inference for Long Context Language Models

When Attention Sink Emerges in Language Models: An Empirical View

Empirical Research on Utilizing LLM-based Agents for Automated Bug Fixing via LangGraph

TableRAG: Million-Token Table Understanding with Language Models

Stress-Testing Long-Context Language Models with Lifelong ICL and Task Haystack

Code Linting using Language Models

Visual Language Model based Cross-modal Semantic Communication Systems

LongSSM: On the Length Extension of State-space Models in Language Modelling