Role-Agent: Bootstrapping LLM Agents via Dual-Role Evolution

要約

Although Large Language Model (LLM) agents have demonstrated strong performance on complex tasks, their learning is often limited by inefficient interaction feedback and static training environments, which hinder broader generalization. To address these limitations, this paper introduces Role-Agent,…