让AI Agent学会发现自身盲点，从而自我进化。
SkillMentor: LLM Agent Self-Evolution via Learning Blind-Spot Diagnosis
- 用强化学习训练一个导师代理，自动诊断执行中的盲点
- 在两个测试环境中平均提升执行性能44.2%
- 无需修改主模型或人工标注，适合自主系统演进
Agent自进化主要关注如何行动，却忽略了发现自身未知同样重要。现有方法通常假设失败可被识别，重点在于修复。本文提出是否能学习‘发现盲点’本身？我们将其视为独立于执行的能力，并排除执行器更新与人类监督两种进步来源。在此约束下，所有性能提升必须来自诊断能力的积累。提出SkillMentor：通过强化学习训练导师策略，生成诊断任务，识别重复失败模式，并整理为可复用的修正技能。在AppWorld和BFCLv3上，技能执行器性能平均提升44.2%。结果表明，盲点诊断是可学习能力，支持无需更新执行器权重或人工数据的自进化。
原文摘要 · Abstract (English)
Agent self-evolution has primarily focused on learning how to act, while overlooking an equally important capability: learning to discover what an agent does not know. Existing approaches typically assume that failure discovery is given, focusing on how to repair failures once they are identified. We ask whether blind-spot diagnosis itself can be learned. We thus study diagnosis as an agent capability separate from execution, and exclude two alternative sources of progress: executor adaptation and human supervision. Under these constraints, performance cannot improve through executor updates or annotated examples, forcing all improvements to originate from the learned diagnostic capability. We propose SkillMentor, which trains a Mentor policy via reinforcement learning to generate diagnostic tasks, identify recurrent failure modes, and curate them into reusable corrective skills. Across AppWorld and BFCLv3, SkillMentor improves executor performance by an average of 44.2%. These results suggest that blind-spot diagnosis is a learnable capability, enabling self-evolution without updating executor weights or relying on human-curated data.
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。