Luke Bailey(@LukeBailey181):Self-play led to superhuman Go performance, why hasn’t it for LLMs? In practice, long run self-play plateaus like RL. We study why this happens, and build a self-play algorithm that scales better. It solves as many problems with a 7B model as the pass@4 of a model 100x bigger.

Luke Bailey

@LukeBailey181

CS PhD student @Stanford advised by @tengyuma & @tatsu_hashimoto. Former CS and Math undergraduate @Harvard. Website:

가입 July 2023

357 팔로잉 중 994 팬

Luke Bailey@LukeBailey181

2026.04.23 15:42

Self-play led to superhuman Go performance, why hasn’t it for LLMs? In practice, long run self-play plateaus like RL. We study why this happens, and build a self-play algorithm that scales better. It solves as many problems with a 7B model as the pass@4 of a model 100x bigger.