HuggingFace Daily Papers·· 1 天前
Agent-Reflex：让语言智能体学会反思性行动
Agent-Reflex: Teaching Language Agents to Act Reflectively
AI 导读
研究者提出 Agent-Reflex，一个激励智能体主动反思的通用训练框架，包含环境反思技能发现（RSD）、反思监督微调（ReSFT）与在策略反思策略优化（ORPO）三阶段。基于 Qwen3-8B 构建的 Agent-Reflex 在五个长程工具使用基准上取得 58.6 总体成绩，在 AgentNoiseBench 上达到 29.1，超过环境扩展与反思基线。代码已开源。
正文
Authors:Guanting Dong, Xinyu Yang, Jiajie Jin, Yuyang Hu, Xiaoshuai Song, Yifei Chen, Xiaoxi Li, Jiejun Tan, Yutao Zhu, Tat-Seng Chua, Ji-Rong Wen, Zhicheng Dou
Abstract:Large language models have evolved from single-turn chatbots into interactive agents capable of long-horizon reasoning. In real deployments, MCP tool services can return noisy feedback and users may intervene to correct the agent, but existing methods provide limited supervision for reflection learning and failure recovery. To bridge this gap, we propose Agent-Reflex, a general training framework for incentivizing proactive reflection in agents. Specifically, Agent-Reflex comprises three stages. First, Environment Reflection Skill Discovery (RSD) injects grounded tool-feedback perturbations into environment scaling, samples a capability-diverse LLM pool, mines matched success-failure pairs, and summarizes reusable reflection meta-skills. Building on this skill library, we then apply Reflective Supervised Fine-Tuning (ReSFT) that converts each skill into supervision for self-reflection, skill recovery, and user-centric correction, strengthening complementary reflective behaviors. Finally, we propose On-Policy Reflective Policy Optimization (ORPO) to combine agentic reinforcement learning under noisy interactions with reflection-guided on-policy distillation to further elicit proactive reflection. Built on Qwen3-8B, Agent-Reflex reaches 58.6 overall on five long-horizon tool-use benchmarks and 29.1 on AgentNoiseBench, surpassing strong environment-scaling and reflection baselines. Experiments on six challenging agent benchmark suites demonstrate the effectiveness of Agent-Reflex. Further analyses show that Agent-Reflex improves agents' robustness to noise while maintaining stable training dynamics, supporting scalable agent training. Our code is available at this https URL.
| Comments: | Working in progress |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.17953 [cs.CL] |
| (or arXiv:2610.17953v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.17953 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Guanting Dong [view email]
[v1]
Tue, 13 Oct 2026 08:36:20 UTC (1,618 KB)
来源：HuggingFace Daily Papers · arxiv.org