Paper 2508.13167
Chain-of-Agents: End-to-End Agent Foundation Models via Multi-Agent Distillation and Agentic RL
- Published
- Aug 2025
- Research lab
- Independent
- Citations
- 66
- GitHub
- 580 stars
01 In brief
Summary
The paper introduces Chain-of-Agents (CoA), a paradigm enabling a single LLM to perform multi-agent problem-solving end-to-end by dynamically activating role-playing and tool agents.
To train these Agent Foundation Models (AFMs), the authors propose multi-agent distillation, which converts trajectories from state-of-the-art multi-agent systems (e.g., OAgents) into CoA format for supervised fine-tuning, followed by agentic reinforcement learning on verifiable tasks.
AFMs achieve state-of-the-art results on web agent benchmarks (GAIA 55.3%, BrowseComp 11.1%, HLE 18.0%) and code/math benchmarks (AIME25 59.8%, LiveCodeBench v5 47.9%), outperforming existing tool-integrated reasoning methods.
The model also reduces inference token consumption by 84.6% compared to traditional multi-agent systems.
The work is fully open-sourced, including model weights, code, and training data, providing a foundation for future research on agent models and agentic RL.
The paper details the CoA paradigm, training framework, experimental setups, and analyses of computational efficiency, generalization to unseen agents, and test-time scaling, demonstrating consistent improvements across diverse benchmarks and model scales (7B and 32B).
02 From the paper
Abstract
Recent advances in large language models (LLMs) and multi-agent systems have demonstrated remarkable capabilities in complex problem-solving tasks such as deep research, vibe coding, and mathematical reasoning. However, most existing multi-agent systems are built upon manual prompt/workflow engineering with sophisticated agent frameworks, making them computationally inefficient, less capable, and can not benefit from data-centric learning. In this work, we introduce Chain-of-Agents (CoA), a novel paradigm of LLM reasoning that enables native end-to-end complex problem-solving in the same way as a multi-agent system (i.e., multi-turn problem solving with multiple tools and multiple agents) within one model. In chain-of-agents problem-solving, the model dynamically activates different tool agents and role-playing agents to simulate multi-agent collaboration in an end-to-end fashion. To elicit end-to-end chain-of-agents problem-solving abilities in LLMs, we introduce a multi-agent distillation framework to distill state-of-the-art multi-agent systems into chain-of-agents trajectories for agentic supervised fine-tuning. We then use agentic reinforcement learning on verifiable agentic tasks to further improve the models' capabilities on chain-of-agents problem solving. We call the resulting models Agent Foundation Models (AFMs). Our empirical studies demonstrate that AFM establishes new state-of-the-art performance across diverse benchmarks in both web agent and code agent settings. We make the entire research, including the model weights, code for training and evaluation, and the training data, fully open-sourced, which offers a solid starting point for future research on agent models and agentic RL.