SOUP: Token-level Single-sample Mix-policy Reinforcement Learning for Large Language Models

Devs

SOUP: Token-level Single-sample Mix-policy Reinforcement Learning for Large Language Models | Read Paper on Bytez