#group-relative advantage

Group-Relative Advantage: 1 AI articles covering group-relative advantage news, analysis, and research

Articles

Li²SAO: Lifting Light Single-Rollout Asynchronous RL

Li²SAO: Lifting Light Single-Rollout Asynchronous RLNEW⭐10

📅 2小时前👁 12 views

Li²SAO improves lightweight single-rollout asynchronous RL for LLMs via group-relative actor-critic bootstrapping, matching GRPO early and raising the later per...