[2603.09117] Decoupling Reasoning and Confidence: Resurrecting Calibration in Reinforcement Learning from Verifiable Rewards

Computer Science > Machine Learning arXiv:2603.09117 (cs) [Submitted on 10 Mar 2026 (v1), last revised 30 Apr 2026 (this version, v2)] Title:Decoupling Reasoning and Confidence: Resurrecting Calibration in Reinforcement Learning from Verifiable Rewards Authors:Zhengzhao Ma, Xueru Wen, Boxi Cao, Yaojie Lu, Hongyu Lin, Jinglin Yang, Min He, Xianpei Han, Le Sun View a PDF of the paper titled Decoupling Reasoning and Confidence: Resurrecting Calibration in Reinforcement Learning from Verifiable Rewards, by Zhengzhao Ma and 8 other authors View PDF HTML (experimental) Abstract:Reinforcement Learning from Verifiable Rewards (RLVR) significantly enhances large language models (LLMs) reasoning but severely suffers from calibration degeneration, where models become excessively over-confident in incorrect answers. Previous studies devote to directly incorporating calibration objective into existing optimization target. However, our theoretical analysis demonstrates that there exists a fundamental gradient conflict between the optimization for maximizing policy accuracy and minimizing calibration error. Building on this insight, we propose DCPO, a simple yet effective framework that systematically decouples reasoning and calibration objectives. Extensive experiments demonstrate that our DCPO not only preserves accuracy on par with GRPO but also achieves the best calibration performance and substantially mitigates the over-confidence issue. Our study provides valuable insights and pra...