Bole Ma

Truncated Proximal Policy Optimization

Model Merging in Pre-training of Large Language Models

DAPO: An Open-Source LLM Reinforcement Learning System at Scale