LLM-based multi-agent systems decompose complex tasks into interacting roles, but most remain manually orchestrated by prompts, tools, and control rules, while agents are rarely optimized through a unified reinforcement learning interface. Existing RL post-training frameworks mainly target single-policy optimization and lack abstractions for user-defined multi-agent workflows, structured interaction, role-specific credit assignment, and configurable parameter sharing.
We present UnityMAS-O, a general RL optimization framework for LLM-based multi-agent systems. UnityMAS-O treats the complete workflow as the optimization unit, rather than a single response or policy trajectory. It represents workflows through four first-class objects: logical agent roles, graph trajectories, user-defined rewards, and agent--model mappings. This decouples logical agents from physical model parameters, supporting full sharing, full separation, and partial sharing, with rewards assigned at role, turn, and trajectory levels.
UnityMAS-O extends verl with a Ray-based star-topology runtime. A central controller executes workflows, invokes tools, records structured trajectories, and assembles rewards; model-local worker groups handle rollout, buffering, advantage computation, and distributed PPO-style updates. Users can define agents, workflows, model mappings, and rewards without rewriting the optimization infrastructure.
We instantiate UnityMAS-O on retrieval-augmented QA, iterative agentic search, and reflective code generation. Across Natural Questions, HotpotQA, and held-out code tasks, multi-agent RL improves manually specified workflows after optimization, with especially large gains for smaller models and strict code all-passed metrics. These results show that UnityMAS-O can serve as a reusable substrate for converting diverse LLM-based multi-agent workflows into trainable multi-agent RL systems.
LLM Agents Factory: Retrieval of Domain-Specific LLM Agents
Sber AI;Moscow Institute of Physics and Technology(莫斯科物理技术学院);National University of Science and Technology MISIS(莫斯科国立科学技术大学MISIS);Skolkovo Institute of Science and Technology(斯科尔科沃科学技术研究所);Artificial Intelligence Research Institute(人工智能研究所)
Department of Computer Science, University College London(计算机科学系,伦敦大学学院);Centre for Artificial Intelligence, University College London(人工智能中心,伦敦大学学院);Department of Computer Science, University of Bologna(计算机科学系,博洛尼亚大学)
Molecular LLM Agents: From Architectural Design to Scientific Autonomy
The Hong Kong Polytechnic University(香港理工大学);Shanghai AI Lab(上海人工智能实验室);National University of Singapore(新加坡国立大学);Shanghai Jiao Tong University(上海交通大学)
AgentBank: Towards Generalized LLM Agents via Fine-Tuning on 50000+ Interaction Trajectories
National Key Laboratory for Multimedia Information Processing(多媒体信息处理国家重点实验室);School of Computer Science, Peking University(北京大学计算机学院);University of Edinburgh(爱丁堡大学);Huawei Technologies(华为技术有限公司)
AgentHarm: A Benchmark for Measuring Harmfulness of LLM Agents
Gray Swan AI(格雷天鹅人工智能公司);UK AI Security Institute(英国人工智能安全研究所);EPFL(洛桑联邦理工学院);Center for AI Safety(人工智能安全中心);Carnegie Mellon University(卡内基梅隆大学);University of Oxford(牛津大学)
From LLMs to LLM-based Agents for Software Engineering: A Survey of Current, Challenges and Future
School of Electrical and Computer Engineering, The University of Sydney(悉尼大学电气与计算机工程学院);University at Buffalo, SUNY(纽约州立大学布法罗分校);University of Chicago(芝加哥大学)
AgentBoard: An Analytical Evaluation Board of Multi-turn LLM Agents
Tsinghua University(清华大学);Shanghai Jiao Tong University(上海交通大学);Westlake University(西湖大学);The University of Hong Kong(香港大学);HKUST(香港科技大学);Zhejiang University(浙江大学)
LLMArena: Assessing Capabilities of Large Language Models in Dynamic Multi-Agent Environments
Beijing University of Posts and Telecommunications(北京邮电大学);Paradigm Inc.(第四范式公司);Tsinghua University(清华大学);The Hong Kong University of Science and Technology (Guangzhou)(香港科技大学(广州))
Watch Out for Your Agents! Investigating Backdoor Threats to LLM-Based Agents
Gaoling School of Artificial Intelligence, Renmin University of China(中国人民大学高瓴人工智能学院);Center for Data Science, Peking University(北京大学数据科学中心);Pattern Recognition Center, WeChat AI, Tencent Inc.(腾讯公司微信AI模式识别中心)
BadAgent: Inserting and Activating Backdoor Attacks in LLM Agents
Zhengzhou University(郑州大学);Institute of Automation, Chinese Academy of Sciences(中国科学院自动化研究所);School of Artificial Intelligence, University of Chinese Academy of Sciences(中国科学院大学人工智能学院)
LLM-Based Agent Society Investigation: Collaboration and Confrontation in Avalon Gameplay
Singapore Management University(新加坡管理大学);Verily Life Sciences(维里生命科学公司);Tencent(腾讯);The Hong Kong University of Science and Technology (Guangzhou)(香港科技大学(广州));The Hong Kong University of Science and Technology(香港科技大学);Singapore University of Technology and Design(新加坡科技设计大学)
LLM Agent Meets Agentic AI: Can LLM Agents Simulate Customers to Evaluate Agentic-AI-based Shopping Assistants?
University of California San Diego La Jolla California United States;Northeastern University Boston Massachusetts United States;North Carolina State University Raleigh North Carolina United States;Independent Researcher Seattle Washington United States;Independent Researcher Palo Alto California United States;University of California San Diego;Northeastern University;North Carolina State University;Independent Researcher