1 min read

slime

Table of Contents

slime is an LLM post-training framework for RL Scaling, developed by THUDM. It provides infrastructure for training large language models with reinforcement learning at scale, supporting algorithms like PPO and GRPO.