GPG: A Simple and Strong Reinforcement Learning Baseline for Model Reasoning
This work addresses the need for efficient reinforcement learning methods to improve model reasoning, though it appears incremental as it builds on existing policy gradient mechanisms.
The paper tackles the problem of enhancing reasoning capabilities in large language models by proposing Group Policy Gradient (GPG), a minimalist reinforcement learning approach that simplifies training by eliminating critic and reference models, and it outperforms GRPO across various tasks while reducing computational costs.
Reinforcement Learning (RL) can directly enhance the reasoning capabilities of large language models without extensive reliance on Supervised Fine-Tuning (SFT). In this work, we revisit the traditional Policy Gradient (PG) mechanism and propose a minimalist RL approach termed Group Policy Gradient (GPG). Unlike conventional methods, GPG directly optimize the original RL objective, thus obviating the need for surrogate loss functions. By eliminating the critic and reference models, avoiding KL divergence constraints, and addressing the advantage and gradient estimation bias, our approach significantly simplifies the training process compared to Group Relative Policy Optimization (GRPO). Our approach achieves superior performance without relying on auxiliary techniques or adjustments. As illustrated in Figure 1, extensive experiments demonstrate that our method not only reduces computational costs but also consistently outperforms GRPO across various unimodal and multimodal tasks. Our code is available at https://github.com/AMAP-ML/GPG.