Agent models: Internalizing Chain-of-Action Generation into Reasoning models
This addresses the need for more autonomous reasoning models in AI, though it appears incremental as it builds on existing methods like ReAct.
The paper tackles the problem of limited autonomy in traditional agentic workflows by proposing Large Agent Models (LAMs) that internalize Chain-of-Action (CoA) generation, resulting in significantly outperforming ReAct-based workflows in task completion on open-domain QA tasks, especially for long-term reasoning and multi-step actions.
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