Off-Policy Correction for LLM RL
Methods for detecting and correcting off-policy drift in reinforcement learning for language models to improve training stability and efficiency.
3 papers
Where this stands
The written synthesis of this thread is for subscribers. Subscribe.