Adaptive Group Policy Optimization

New RL methods tackle LLM training issues

Two new research papers introduce methods to improve the training of large language models using reinforcement learning. One paper addresses the issue of "advantage collapse" in Group Relative Policy Optimization (GRPO)…