GDPO: Group reward-Decoupled Normalization Policy Optimization for Multi-reward RL Optimization

Shih-Yang Liu; Xin Dong; Ximing Lu; Shizhe Diao; Peter Belcak; Mingjie Liu; Min-Hung Chen; Hongxu Yin; Yu-Chiang Frank Wang; Kwang-Ting Cheng; Yejin Choi; Jan Kautz; Pavlo Molchanov

Paper · arXiv 2601.05242

GDPO: Group reward-Decoupled Normalization Policy Optimization for Multi-reward RL Optimization

Shih-Yang Liu, Xin Dong, Ximing Lu, Shizhe Diao, Peter Belcak, Mingjie Liu, Min-Hung Chen, Hongxu Yin, Yu-Chiang Frank Wang, Kwang-Ting Cheng, Yejin Choi, Jan Kautz, Pavlo Molchanov

235 upvotesJanuary 8, 2026arXiv 预印本

AI 摘要

Multi-reward reinforcement learning suffers from reward normalization collapse in GRPO, which GDPO addresses by decoupling reward normalization for improved training stability and performance across reasoning tasks.

Reinforcement learningGroup Relative Policy Optimizationmulti-reward settingpolicy optimizationGroup reward-Decoupled Normalization Policy Optimizationreward normalizationadvantage valuestraining stabilitymulti-reward reinforcement learning

Abstract

As language models become increasingly capable, users expect them to provide not only accurate responses but also behaviors aligned with diverse human preferences across a variety of scenarios. To achieve this, Reinforcement learning (RL) pipelines have begun incorporating multiple rewards, each capturing a distinct preference, to guide models toward these desired behaviors. However, recent work has defaulted to apply Group Relative Policy Optimization (GRPO) under multi-reward setting without examining its suitability. In this paper, we demonstrate that directly applying GRPO to normalize distinct rollout reward combinations causes them to collapse into identical advantage values, reducing the resolution of the training signal and resulting in suboptimal convergence and, in some cases, early training failure. We then introduce Group reward-Decoupled Normalization Policy Optimization (GDPO), a new policy optimization method to resolve these issues by decoupling the normalization of individual rewards, more faithfully preserving their relative differences and enabling more accurate multi-reward optimization, along with substantially improved training stability. We compare GDPO with GRPO across three tasks: tool calling, math reasoning, and coding reasoning, evaluating both correctness metrics (accuracy, bug ratio) and constraint adherence metrics (format, length). Across all settings, GDPO consistently outperforms GRPO, demonstrating its effectiveness and generalizability for multi-reward reinforcement learning optimization.