Group Sequence Policy Optimization
By: Chujie Zheng , Shixuan Liu , Mingze Li and more
Potential Business Impact:
Teaches AI to learn better and faster.
This paper introduces Group Sequence Policy Optimization (GSPO), our stable, efficient, and performant reinforcement learning algorithm for training large language models. Unlike previous algorithms that adopt token-level importance ratios, GSPO defines the importance ratio based on sequence likelihood and performs sequence-level clipping, rewarding, and optimization. We demonstrate that GSPO achieves superior training efficiency and performance compared to the GRPO algorithm, notably stabilizes Mixture-of-Experts (MoE) RL training, and has the potential for simplifying the design of RL infrastructure. These merits of GSPO have contributed to the remarkable improvements in the latest Qwen3 models.
Similar Papers
Group Sequence Policy Optimization
Machine Learning (CS)
Makes AI learn faster and better.
Agent-GSPO: Communication-Efficient Multi-Agent Systems via Group Sequence Policy Optimization
Multiagent Systems
Makes AI agents talk less, saving money.
Single-stream Policy Optimization
Machine Learning (CS)
Teaches AI to solve harder math problems better.