Self-play limitations refer to the constraints and failure modes of self-play approaches in reinforcement learning and model distillation. While self-play has been highly successful in games (e.g., AlphaGo), its application to LLM post-training and distillation faces challenges including reward hacking, distribution collapse, and the lack of a stable ground-...
- Related
- References
Self-Play Limitations
Self-play limitations refer to the constraints and failure modes of self-play approaches in reinforcement learning and model distillation. While self-play has been highly successful in games (e.g., AlphaGo), its application to LLM post-training and distillation faces challenges including reward hacking, distribution collapse, and the lack of a stable ground-truth objective. The on-policy-distillation-survey discusses these limitations in the context of modern LLM training.
Related
- on-policy-distillation-survey — Surveys limitations of self-play in distillation contexts
- on-policy-self-distillation — Self-distillation as an alternative to traditional self-play