mlsanityData to 14 Oct 2026

Agent-Reflex: Teaching Language Agents to Act Reflectively

arXiv 2610.17953 ↗First posted 13 Oct 2026cs.CLcited by 0

On the map: LLM agents and tool use → Reinforcement learning for LLM agents

Abstract (arXiv)

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 https://github.com/RUC-NLPIR/Agent-Reflex.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Reinforcement learning for LLM agents, one of 16 directions in LLM agents and tool use. Builds on 29 of its 42 references that are in the map. First posted on arXiv 13 Oct 2026.

Builds on

29 of 42 references.

Closest ancestors · 2

Same direction, newest first

  1. 2026SEED: Self-Evolving On-Policy Distillation for Agentic Reinforcement Learning16 Jul
  2. 2025Simulating Environments with Reasoning Models for Agent Training3 Nov

Other references · 27

  1. 2024HybridFlow: A Flexible and Efficient RLHF Frameworkcited by 2,057
  2. 2023Improving Factuality and Reasoning in Language Models through Multiagent Debatecited by 1,784
  3. 2024LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Modelscited by 1,464
  4. 2023On-Policy Distillation of Language Models: Learning from Self-Generated Mistakescited by 939
  5. 2023ExpeL: LLM Agents Are Experiential Learnerscited by 876
  6. 2025τ2-Bench: Evaluating Conversational Agents in a Dual-Control Environmentcited by 627
  7. 2023LLM-Blender: Ensembling Large Language Models with Pairwise Ranking and Generative Fusioncited by 564
  8. 2026Self-Distilled Reasoner: On-Policy Self-Distillation for Large Language Modelscited by 455
  9. 2026SkillsBench: Benchmarking How Well Agent Skills Work Across Diverse Taskscited by 315
  10. 2024ToolSandbox: A Stateful, Conversational, Interactive Evaluation Benchmark for LLM Tool Use Capabilitiescited by 295
  11. 2025REINFORCE++: Stabilizing Critic-Free Policy Optimization with Global Advantage Normalizationcited by 241
  12. 2025Tool-Star: Empowering LLM-Brained Multi-Tool Reasoner via Reinforcement Learningcited by 72
15 more
  1. 2025ACEBench: Who Wins the Match Point in Tool Usage?cited by 71
  2. 2025TOUCAN: Synthesizing 1.5M Tool-Agentic Data from Real-World MCP Environmentscited by 69
  3. 2026Agent World Model: Infinity Synthetic Environments for Agentic Reinforcement Learningcited by 66
  4. 2026EnvScaler: Scaling Tool-Interactive Environments for LLM Agent via Programmatic Synthesiscited by 63
  5. 2026Agent-World: Scaling Real-World Environment Synthesis for Evolving General Agent Intelligencecited by 54
  6. 2025VitaBench: Benchmarking LLM Agents with Versatile Interactive Tasks in Real-world Applicationscited by 53
  7. 2025GenEnv: Difficulty-Aligned Co-Evolution Between LLM Agents and Environment Simulatorscited by 42
  8. 2026ReliabilityBench: Evaluating LLM Agent Reliability Under Production-Like Stress Conditionscited by 36
  9. 2026ScaleEnv: Scaling Environment Synthesis from Scratch for Generalist Interactive Tool-Use Agent Trainingcited by 32
  10. 2026Workspace-Bench 1.0: Benchmarking AI Agents on Workspace Tasks with Large-Scale File Dependenciescited by 29
  11. 2026AgentNoiseBench: Benchmarking Robustness of Tool-Using LLM Agents Under Noisy Conditioncited by 23
  12. 2022Continual Learning for Instruction Following from Realtime Feedbackcited by 22
  13. 2026Learning to Act under Noise: Enhancing Agent Robustness via Noisy Environmentscited by 6
  14. 2026ReflectRL: Learning from Golden Negative Trajectories via Reflective-to-Direct Reasoningcited by 6
  15. 2026LoongReflect: Boosting Long-Horizon Reflection in Search Agents via Global Perspective Distillationcited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2023Reflexion: Language Agents with Verbal Reinforcement Learning
  2. 2026Agentic Critical Training
  3. 2026Internalizing Agency from Reflective Experience
  4. 2024MetaReflection: Learning Instructions for Language Agents using Past Reflections
  5. 2026Experiential Reflective Learning for Self-Improving LLM Agents
  6. 2026Learn Like Humans: Use Meta-cognitive Reflection for Efficient Self-Improvement
  7. 2025Memento 2: Learning by Stateful Reflective Memory
  8. 2026Experiential Reinforcement Learning