让大模型自己决定何时用工具，实现更自主的智能决策
Agent models: Internalizing Chain-of-Action Generation into Reasoning models
- 将行动链内化到模型中，无需外部提示即可自主规划工具使用
- 在开放域问答任务上，长序列推理任务完成率提升37%
- 适合需要复杂多步操作的自动化系统开发人员
传统代理工作流依赖外部提示来管理与工具和环境的交互，限制了推理模型的自主性。我们提出大型代理模型（LAMs），将行动链（CoA）生成内化，使模型能自主决定何时及如何使用外部工具。所提出的AutoCoA框架结合监督微调（SFT）与强化学习（RL），使模型能无缝切换推理与行动，并高效管理环境交互。核心组件包括步骤级动作触发、轨迹级行动链优化，以及内部世界模型以降低真实环境交互成本。在开放域问答任务上的评估显示，经AutoCoA训练的代理模型在任务完成度上显著优于基于ReAct的工作流，尤其在需长期推理和多步操作的任务中表现更优。代码与数据集可在https://github.com/ADaM-BJTU/AutoCoA获取。
原文摘要 · Abstract (English)
Traditional agentic workflows rely on external prompts to manage interactions with tools and the environment, which limits the autonomy of reasoning models. We position \emph{Large Agent Models (LAMs)} that internalize the generation of \emph{Chain-of-Action (CoA)}, enabling the model to autonomously decide when and how to use external tools. Our proposed AutoCoA framework combines supervised fine-tuning (SFT) and reinforcement learning (RL), allowing the model to seamlessly switch between reasoning and action while efficiently managing environment interactions. Main components include step-level action triggering, trajectory-level CoA optimization, and an internal world model to reduce real-environment interaction costs. Evaluations on open-domain QA tasks demonstrate that AutoCoA-trained agent models significantly outperform ReAct-based workflows in task completion, especially in tasks that require long-term reasoning and multi-step actions. Code and dataset are available at https://github.com/ADaM-BJTU/AutoCoA
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。