Is One Layer Enough? Training A Single Transformer Layer Can Match Full-Parameter RL Training

By Zijian Zhang · Paper · cs.LG

Reinforcement learning (RL) has become a central component of post-training large language models (LLMs), yet little is understood about how RL adaptation is distributed across transformer layers. Existing approaches typically update all model parameters uniformly, implicitly ass

Cs.lg

View original

HomeResourceLoading…