Is One Layer Enough? A Single Transformer Layer Matches Full-Parameter RL Train

by tcp_handshaker | View on Hacker News