Large language models are transforming systems research by automating the discovery of performance-critical algorithms for computer systems. Despite plausible codes generated by LLMs, producing solutions that meet the stringent correctness and performance requirements of systems demands iterative optimization. Test-time reinforcement learning offers high search efficiency but requires parameter updates infeasible under API-only access, while existing training-free evolutionary methods suffer from inefficient context utilization and undirected search. We introduce ContextEvolve, a multi-agent framework that achieves RL-level search efficiency under strict parameter-blind constraints by decomposing optimization context into three orthogonal dimensions: a Summarizer Agent condenses semantic state via code-to-language abstraction, a Navigator Agent distills optimization direction from trajectory analysis, and a Sampler Agent curates experience distribution through prioritized exemplar retrieval. This orchestration forms a functional isomorphism with RL-mapping to state representation, policy gradient, and experience replay-enabling principled optimization in a textual latent space. On the ADRS benchmark, ContextEvolve outperforms state-of-the-art baselines by 33.3% while reducing token consumption by 29.0%. Codes for our work are released at https://anonymous.4open.science/r/ContextEvolve-ACC
Reinforcement Learning for Chain of Thought Compression with One-Domain-to-All Generalization
CFCS, School of Computer Science, Peking University(计算机学院,北京大学);State Key Laboratory of Multimedia Information Processing, School of Computer Science, Peking University(多媒体信息处理国家重点实验室,计算机学院,北京大学);LLM-Core Xiaomi(小米LLM-Core)
Self-Compression of Chain-of-Thought via Multi-Agent Reinforcement Learning
Renmin University of China(中国人民大学);Xiaohongshu Inc.(小红书公司);Institute of Automation,Chinese Academy of Sciences(中国科学院自动化研究所);University of Southern California(南加州大学);Shandong University(山东大学)
AOI: Context-Aware Multi-Agent Operations via Dynamic Scheduling and Hierarchical Memory Compression
Columbia University(哥伦比亚大学);Hunan University(湖南大学);Central University of Finance and Economics(中央财经大学);Chongqing University(重庆大学);Institute of Automation, Chinese Academy of Sciences(中国科学院自动化研究所);AI Agent Lab, Vokram Group(Vokram集团人工智能代理实验室);College of William and Mary(威廉与玛丽学院);University of Texas(得克萨斯大学);Zhengzhou University of Light Industry(郑州轻工业大学);Emory University(埃默里大学);Department of Nephrology, Affiliated Hospital of Guangdong Medical University(广东医科大学附属医院肾病科)
OmniDrive: An LLM-Choreographed Multi-Agent World Model with Unified Latent Co-Compression for Multi-View Driving Video Generation
Peking University(北京大学);Xiamen University(厦门大学);Korea Advanced Institute of Science and Technology (KAIST)(韩国科学技术院);National Taiwan University(国立台湾大学);Wuhan University(武汉大学);Wuhan University of Technology(武汉理工大学);Tsinghua University(清华大学);Jimei University(集美大学)
New York University(纽约大学);Modal Labs(Modal实验室);University of Maryland(马里兰大学);Princeton University(普林斯顿大学);Columbia University(哥伦比亚大学);Harvard University(哈佛大学);Lawrence Livermore National Laboratory(劳伦斯利弗莫尔国家实验室);FAIR at Meta(Meta FAIR实验室)
Enabling Agents to Communicate Entirely in Latent Space
State Key Lab of CAD&CG(CAD与CG国家重点实验室);Future Living Lab of Alibaba(阿里巴巴未来生活实验室);Zhejiang Key Laboratory of Medical Imaging Artificial Intelligence(浙江医学影像人工智能重点实验室);Shanghai Jiao Tong University(上海交通大学)
An Information Theoretic Perspective on Agentic System Design
Department of Computer Science, Stanford University(斯坦福大学计算机科学系);Department of Statistics, Stanford University(斯坦福大学统计学系);Wu Tsai Neurosciences Institute, Stanford University(斯坦福大学吴泰教授神经科学研究所)
MAAM: A Lightweight Multi-Agent Aggregation Module for Efficient Image Classification Based on the MindSpore Framework
School of Computing and Information(计算与信息学院);Network Security Research Center(网络安全研究中心);Big Data and Policing Technology Laboratory(大数据与警务技术实验室);Guangxi Police College(广西警察学院);Chinese Academy of Sciences(中国科学院);Institute of Software(软件研究所)