More Methods, Better Agents? A Reality Check on Self-Distillation in Agentic Reinforcement Learning
🔗 https://huggingface.co/papers/2610.19789 ↗Researchers benchmarked ten self-distillation methods against GRPO across multiple agent tasks and model sizes, finding limited consistent gains, hidden training instability, and no reliable cost-to-performance benefit.
內容為收集當日抓取的摘要原文。