More Methods, Better Agents? A Reality Check on Self-Distillation in Agentic Reinforcement Learning

arXiv:2610.19789 · cs.CL, cs.LG · Submitted 2026-10-14 · Read on arXiv

cs.CL, cs.LG

Submitted: 2026-10-14

Updated: 2026-10-14

Code: https://github.com/aim-uofa/Self-Distillation-AgentRL

Project page: https://aim-uofa.github.io/Self-Distillation-AgentRL

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Distillation — 1× in this paper · explained in 2 episodes
• Reinforcement Learning — 1× in this paper · explained in 4 episodes
• Self-Distillation — 1× in this paper · explained in 7 episodes

Sources

• ReST meets ReAct: Self-Improvement for Multi-Step Reasoning LLM Agent
• FireAct: Toward Language Agent Fine-tuning
• Privileged, but Biased: How PI-Conditioned Teachers Break Self-Distillation
• Distilling the Knowledge in a Neural Network
• Unifying distillation and privileged information
• Self-Distilled Agentic Reinforcement Learning
• PCSD: Persistent Consistency for Self-Distillation in Agentic Reinforcement Learning
• Policy Distillation
• DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
• ALFWorld: Aligning Text and Embodied Environments for Interactive Learning
• Skill-SD: Skill-Conditioned Self-Distillation for Multi-turn LLM Agents
• AgentOPSD: Recursive Self-Distillation for Agentic Reinforcement Learning
• RAGEN: Understanding Self-Evolution in LLM Agents via Multi-Turn Reinforcement Learning
• From Scoring to Acting: Outcome-Verified Comparative Self-Distillation for LLM Agents
• Self-Distilled RLVR
• Why Does Feedback-Augmented Self-Distillation Fail to Improve Retrieval-Interleaved Search Agents?
• Agentic Reinforcement Learning with Observation-Calibrated Self-Distillation
• ReAct: Synergizing Reasoning and Acting in Language Models
• Agentic Reinforcement Learning with Self-Distilled Reward Shaping

Related papers

• Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
• Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
• Subliminal Steering: Stronger Encoding of Hidden Signals
• MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
• The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
• Untangling the Mechanisms of Misleading Context in Medical Question Answering