语言大模型 / LLM

LOGO -- Long cOntext aliGnment via efficient preference Optimization

Zecheng Tang, Zechen Sun, Juntao Li, Qiaoming Zhu, Min Zhang

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

Explore, Select, Derive, and Recall: Augmenting LLM with Human-like Memory for Mobile Task Automation

Sunjae Lee, Junyoung Choi, Jungjae Lee, Munim Hasan Wasi, Hojun Choi, Steven Y. Ko, Sangeun Oh, Insik Shin

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

MEOW: MEMOry Supervised LLM Unlearning Via Inverted Facts

Tianle Gu, Kexin Huang, Ruilin Luo, Yuanqi Yao, Yujiu Yang, Yan Teng, Yingchun Wang

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

Untie the Knots: An Efficient Data Augmentation Strategy for Long-Context Pre-Training in Language Models

Junfeng Tian, Da Zheng, Yang Cheng, Rui Wang, Colin Zhang, Debing Zhang

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

Granite Code Models: A Family of Open Foundation Models for Code Intelligence

Mayank Mishra, Matt Stallone, Gaoyuan Zhang, Yikang Shen, Aditya Prasad, Adriana Meza Soria, Michele Merler, Parameswaran Selvam, Saptha Surendran, Shivdeep Singh, Manish Sethi, Xuan-Hong Dang, Pengyuan Li, Kun-Lung Wu, Syed Zawad, Andrew Coleman, Matthew White, Mark Lewis, Raju Pavuluri, Yan Koyfman, Boris Lublinsky, Maximilien de Bayser, Ibrahim Abdelaziz, Kinjal Basu, Mayank Agarwal, Yi Zhou, Chris Johnson, Aanchal Goyal, Hima Patel, Yousaf Shah, Petros Zerfos, Heiko Ludwig, Asim Munawar, Maxwell Crouse, Pavan Kapanipathi, Shweta Salaria, Bob Calio, Sophia Wen, Seetharami Seelam, Brian Belgodere, Carlos Fonseca, Amith Singhee, Nirmit Desai, David D. Cox, Ruchir Puri, Rameswar Panda

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

Comments Corresponding Authors: Rameswar Panda, Ruchir Puri; Equal Contributors: Mayank Mishra, Matt Stallone, Gaoyuan Zhang

Efficient Streaming Language Models with Attention Sinks

Guangxuan Xiao, Yuandong Tian, Beidi Chen, Song Han, Mike Lewis

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

Comments ICLR 2024

DataFrame QA: A Universal LLM Framework on DataFrame Question Answering Without Data Exposure

Junyi Ye, Mengnan Du, Guiling Wang

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

ToolkenGPT: Augmenting Frozen Language Models with Massive Tools via Tool Embeddings

Shibo Hao, Tianyang Liu, Zhen Wang, Zhiting Hu

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

Comments NeurIPS 2023 (oral). Code: https://github.com/Ber666/ToolkenGPT

FinMem: A Performance-Enhanced LLM Trading Agent with Layered Memory and Character Design

Yangyang Yu, Haohang Li, Zhi Chen, Yuechen Jiang, Yang Li, Denghui Zhang, Rong Liu, Jordan W. Suchow, Khaldoun Khashanah

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

Contextual Memory Virtualisation: DAG-Based State Management and Structurally Lossless Trimming for LLM Agents

上下文记忆虚拟化：基于DAG的状态管理和结构无损修剪用于LLM代理

Cosmo Santoni

机构 * Imperial College London（伦敦帝国学院） 纠错

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

AI总结 上下文记忆虚拟化通过DAG结构管理和无损修剪技术，提升LLM代理在长期推理任务中的上下文重用效率和经济性。

Comments 11 pages. 6 figures. Introduces a DAG-based state management system for LLM agents. Evaluation on 76 coding sessions shows up to 86% token reduction (mean 20%) while remaining economically viable under prompt caching. Includes reference implementation for Claude Code

SACTOR: LLM-Driven Correct and Idiomatic C to Rust Translation with Static Analysis and FFI-Based Verification

SACTOR：基于大语言模型的C到Rust翻译工具，结合静态分析和FFI验证

Tianyang Zhou, Ziyi Zhang, Haowen Lin, Somesh Jha, Mihai Christodorescu, Kirill Levchenko, Varun Chandrasekaran

机构 * University of Illinois Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校） ； University of Wisconsin–Madison（威斯康星大学麦迪逊分校） ； Google（谷歌） 纠错

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

AI总结 SACTOR通过结合静态分析和FFI验证，利用大语言模型实现C到Rust的正确且习惯性翻译，提升了代码安全性和效率

Comments 35 pages, 15 figures Previously named as "LLM-Driven Multi-step Translation from C to Rust using Static Analysis"

Context Engineering for Multi-Agent LLM Code Assistants Using Elicit, NotebookLM, ChatGPT, and Claude Code

Muhammad Haseeb

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

Comments 15 pages, 5 figures, research paper on multi-agent LLM systems for code generation

FinCon: A Synthesized LLM Multi-Agent System with Conceptual Verbal Reinforcement for Enhanced Financial Decision Making

Yangyang Yu, Zhiyuan Yao, Haohang Li, Zhiyang Deng, Yupeng Cao, Zhi Chen, Jordan W. Suchow, Rong Liu, Zhenyu Cui, Zhaozhuo Xu, Denghui Zhang, Koduvayur Subbalakshmi, Guojun Xiong, Yueru He, Jimin Huang, Dong Li, Qianqian Xie

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

Comments LLM Applications, LLM Agents, Financial Technology, Quantitative Finance, Algorithmic Trading, Cognitive Science

AutoDroid: LLM-powered Task Automation in Android

Hao Wen, Yuanchun Li, Guohong Liu, Shanhui Zhao, Tao Yu, Toby Jia-Jun Li, Shiqi Jiang, Yunhao Liu, Yaqin Zhang, Yunxin Liu

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

Comments Published in MobiCom 2024; Original title: "Empowering LLM to use Smartphone for Intelligent Task Automation"

WebChoreArena: Evaluating Web Browsing Agents on Realistic Tedious Web Tasks

WebChoreArena：在现实繁琐网页任务上评估网页浏览智能体

Atsuyuki Miyai, Zaiying Zhao, Kazuki Egashira, Atsuki Sato, Tatsumi Sunada, Shota Onohara, Hiromasa Yamanishi, Mashiro Toyooka, Kunato Nishina, Ryoma Maeda, Kiyoharu Aizawa, Toshihiko Yamasaki

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

AI总结 本文提出WebChoreArena，作为WebArena的扩展，包含532个耗时300多小时开发的繁琐网页任务，从大规模记忆、计算、长期记忆三维度评估网页浏览智能体，实验显示其能清晰衡量LLM进展且难度高于WebArena。

Comments COLM2026. Project Page: https://webchorearena.github.io/

RUTA: Principled Visual Token Allocation via Rate-Utility Optimization

RUTA：基于率-效用优化的原则性视觉令牌分配方法

Jian Zou, Xiaoyu Xu, Zhihua Wang, Yilin Wang, Balu Adsumilli, Kede Ma

机构 * City University of Hong Kong（香港城市大学） ； Google Inc.（谷歌公司） 纠错

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

AI总结 针对视觉语言模型因长视觉令牌序列导致的LLM侧计算和内存成本过高问题，提出RUTA方法，仅用2.0%、4.2%的视觉令牌，分别在LLaVA-NeXT-7B、Qwen3-VL-8B上保留88.2%、94.4%的任务性能。

Formal Verification of an Out-of-Order Multiprocessor against an In-Order Weak-Memory ISA

针对顺序弱内存ISA的乱序多处理器形式验证

Janggun Lee, Jeehoon Kang

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

AI总结 研究针对顺序弱内存ISA的乱序多处理器验证问题，核心方法是设计核心规范并分两步证明，主要贡献是首次实现此类形式验证，借助核心规范简化证明，且利用LLM代理自动编写证明。

From Blind Search to Memory-Aware Evolution: Efficient DBMS Tuning via Collaborative Diagnosis and Utility-Aware Retrieval

从盲目搜索到内存感知进化：通过协作诊断和效用感知检索实现高效数据库管理系统调优

Zhaoyan Hong, Yishen Sun, Xinyi Zhang, Zhentao Han, Jinhao Dong, Wei Lu, Kai Xu, Liu Tang, Qi Liu, Xiaoyong Du

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

AI总结 针对多组件DBMS调优难题，EvoTune框架通过协作诊断定位高影响子空间，引入效用感知检索策略，将调优反馈组织成层次化内存，无需LLM微调，实验证明其性能优于现有基线，能快速提升查询性能。

General Non-Clairvoyant KV-Cache Scheduling via Regime-Aware Routing

通过状态感知路由实现通用的非先知式键值缓存调度

Yiding Feng, Siyu Liu, Zonghan Yang, Yuhao Zhang

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

AI总结 研究硬KV缓存内存预算下批量LLM推理的非先知式调度，提出基于状态感知路由框架的常数竞争算法，能处理任意提示和响应长度，对完工时间和总完成时间有常数竞争保证。

Seeing and Reflecting: Multimodal Memory-Enhanced Agent Collaboration for Recommendation

视觉与反思：用于推荐的多模态记忆增强智能体协作

Hao Cong, Huizu Lin, Zihan Wang, Chengkai Huang, Quan Z. Sheng, Lina Yao

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

AI总结 针对基于LLM的推荐系统局限，提出MMEACR框架，采用双轨记忆架构，通过属性引导机制更新记忆，构建多模态嵌入记忆，经加权倒数排名融合两轨道，实验证明其在多领域表现出色，尤其视觉推荐场景有显著提升。

ConCovUp: Effective Agent-Based Test Driver Generation for Concurrency Testing

ConCovUp：面向并发测试的有效代理基于测试驱动生成

Yuandao Cai, Shuhao Fu, Wensheng Tang, Cheng Wen, Shengchao Qin, Charles Zhang

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

AI总结 本文提出ConCovUp框架，结合LLM与程序分析，通过静态分析提取共享内存访问及调用上下文，利用LLM驱动的逆向追踪方法生成有效并发测试，提升SMAP覆盖率。

Illusion-Aware Visual Preprocessing and Anti-Illusion Prompting for Classic Illusion Understanding in Vision-Language Models

基于 illusion 意识的视觉预处理与反 illusion 提示的经典 illusion 理解在视觉-语言模型中的应用

Junli Zha, Jiahui Wang, Xinkai Lu, Jinbo Wang

机构 * SF Technology Co., Ltd.（SF技术有限公司） 纠错

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

AI总结 本文提出无需训练的框架，通过图像预处理、反 illusion 提示工程和多票集成策略，解决视觉-语言模型对视觉错觉的感知与记忆冲突，实现90.48%和98.41%的准确率。

Comments Accepted at CVPR 2026 Workshop on 5th DataCV Challenge

FlashRT: Towards Computationally and Memory Efficient Red-Teaming for Prompt Injection and Knowledge Corruption

FlashRT: 向计算和内存高效方向发展用于提示注入和知识腐败的红队测试

Yanting Wang, Chenlong Yin, Ying Chen, Jinyuan Jia

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

AI总结 本文提出FlashRT框架，通过提升计算和内存效率，优化长上下文LLM的提示注入和知识腐败攻击，实现2-7倍的加速和2-4倍的内存节省，为系统评估长上下文LLM的安全性提供工具。

Comments The code is available at https://github.com/Wang-Yanting/FlashRT

SAGE: Selective Attention-Guided Extraction for Token-Efficient Document Indexing

SAGE：基于选择性注意力的令牌高效文档索引提取

Xinzhi Wang, Peter Baile Chen, Gerardo Vitagliano, Matthew Russo, Jun Chen, Michael Cafarella, Samuel Madden, Chunwei Liu

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

AI总结 本文提出SAGE方法，通过轻量本地LLM进行单次预填充，将语言模型注意力信号转化为查询特定的相关热图，从而在可控粒度下选择高相关单元，减少上下文并提升问答性能。

Comments 12 pages, 10 figures

Search-Based Multi-Trajectory Refinement for Safe C-to-Rust Translation with Large Language Models

基于搜索的多轨迹优化用于大型语言模型驱动的安全C到Rust转换

HoHyun Sim, Hyeonjoong Cho, Yeonghyeon Go, Sadegh AlMahdi Kazemi Zarkouei, Zhoulai Fu, Ali Shokri, Binoy Ravindran

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

AI总结 本文提出LAC2R方法，利用MCTS系统探索多轨迹优化，提升C到Rust转换的安全性和准确性，实验表明其在大规模和小规模基准测试中均表现优异。

LAAF: Logic-layer Automated Attack Framework A Systematic Red-Teaming Methodology for LPCI Vulnerabilities in Agentic Large Language Model Systems

LAAF：逻辑层自动攻击框架：一种针对代理大语言模型系统中LPCI漏洞的系统性红队方法

Hammad Atta, Ken Huang, Kyriakos Rock Lambros, Yasir Mehmood, Zeeshan Baig, Mohamed Abdur Rahman, Manish Bhatt, M. Aziz Ul Haq, Muhammad Aatif, Nadeem Shahzad, Kamal Noor, Vineeth Sai Narajala, Hazem Ali, Jamel Abed

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

AI总结 LAAF是首个结合特定技术分类和分阶段种子升级的自动化红队框架，通过49种技术分类生成大量独特负载，并通过持久化阶段突破器实现分阶段负载变异，有效提升阶段突破效率。

ReFusion: A Diffusion Large Language Model with Parallel Autoregressive Decoding

ReFusion：一种具有并行自回归解码的扩散大语言模型

Jia-Nan Li, Jian Guan, Wei Wu, Chongxuan Li

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

AI总结 ReFusion通过整合序列重组到因果注意框架中，实现并行解码和高效生成，显著提升性能和速度，缩小与传统自回归模型的差距。

Eagle 2.5: Boosting Long-Context Post-Training for Frontier Vision-Language Models

Guo Chen, Zhiqi Li, Shihao Wang, Jindong Jiang, Yicheng Liu, Lidong Lu, De-An Huang, Wonmin Byeon, Matthieu Le, Tuomas Rintamaki, Tyler Poon, Max Ehrlich, Tuomas Rintamaki, Tyler Poon, Tong Lu, Limin Wang, Bryan Catanzaro, Jan Kautz, Andrew Tao, Zhiding Yu, Guilin Liu

专题命中 长上下文与记忆 ：language model(title,abstract)；post-training(title)

Agentao: A Governed Local-First Runtime for Tool-Using LLM Agents

Agentao：面向使用工具的大语言模型智能体的受管控本地优先运行时

Bo Jin, Qiang Jiao, Xin Tong

机构 * The Third Research Institute of the Ministry of Public Security（公安部第三研究所） ； Bureau of Science and Technology Information Ministry of Public Security of the People’s Republic of China（中华人民共和国公安部科技信息局） ； School of Information and Cyber Security People’s Public Security University of China（中国人民公安大学信息与网络安全学院） 纠错

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

AI总结 针对工具使用型LLM智能体的安全管控需求，提出Agentao运行时，通过分层架构分离动作提案与授权执行，将权限等构建为显式抽象，提升智能体可管控性与可检查性。

Comments The code is publicly available at Github. We are conducting testing and analysis of this framework, and will provide experimental results and examples in future versions

Interpreting Language Model Hidden States at Scale

大规模语言模型隐藏状态的解释

Jordan Pettyjohn, Mansi Sakarvadia, Nathaniel Hudson, Daniel McKenzie, Kyle Chard, Ian Foster

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

AI总结 研究人员提出OmniLens，一种可适配任意模型宽度激活的Lens方法，通过低秩翻译器和Subset-KL技术降低成本，在LLaMA-3.3-70B上构建482个Lens集成，以更低成本复现提示注入检测等案例研究的关键结果。