Saved in:
Bibliographic Details
Main Authors: Zeng, Xianzhou, Huang, Jing, Xie, Chunmei, Nan, Gongrui, Chen, Siye, Lu, Mengyu, Xiong, Weiqi, Zhou, Qixuan, Zhang, Junhao, Zhu, Qiang, Li, Yadong, Xu, Xingzhong
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2601.22648
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916048557572096
author Zeng, Xianzhou
Huang, Jing
Xie, Chunmei
Nan, Gongrui
Chen, Siye
Lu, Mengyu
Xiong, Weiqi
Zhou, Qixuan
Zhang, Junhao
Zhu, Qiang
Li, Yadong
Xu, Xingzhong
author_facet Zeng, Xianzhou
Huang, Jing
Xie, Chunmei
Nan, Gongrui
Chen, Siye
Lu, Mengyu
Xiong, Weiqi
Zhou, Qixuan
Zhang, Junhao
Zhu, Qiang
Li, Yadong
Xu, Xingzhong
contents The key to building trustworthy large language models (LLMs) lies in endowing them with inherent uncertainty expression capabilities, thereby mitigating overconfident errors in high-stakes applications. However, existing RL paradigms such as GRPO often suffer from Advantage Bias due to binary decision spaces and static uncertainty rewards, inducing either excessive conservatism or overconfidence. To tackle this challenge, this paper unveils the root causes of reward hacking and overconfidence in current RL paradigms incorporating uncertainty-based rewards, based on which we propose the UnCertainty-Aware Policy Optimization (UCPO) framework. UCPO employs Ternary Advantage Decoupling to separate and independently normalize deterministic and uncertain rollouts, thereby eliminating advantage bias. Furthermore, a Dynamic Uncertainty Reward Adjustment mechanism adapts uncertainty weights in real-time according to model evolution and instance difficulty. Experimental results in mathematical reasoning and general tasks demonstrate that UCPO effectively resolves the reward imbalance, significantly improving the reliability of the model beyond their knowledge boundaries.
format Preprint
id arxiv_https___arxiv_org_abs_2601_22648
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle UCPO: Uncertainty-Aware Policy Optimization
Zeng, Xianzhou
Huang, Jing
Xie, Chunmei
Nan, Gongrui
Chen, Siye
Lu, Mengyu
Xiong, Weiqi
Zhou, Qixuan
Zhang, Junhao
Zhu, Qiang
Li, Yadong
Xu, Xingzhong
Artificial Intelligence
Machine Learning
The key to building trustworthy large language models (LLMs) lies in endowing them with inherent uncertainty expression capabilities, thereby mitigating overconfident errors in high-stakes applications. However, existing RL paradigms such as GRPO often suffer from Advantage Bias due to binary decision spaces and static uncertainty rewards, inducing either excessive conservatism or overconfidence. To tackle this challenge, this paper unveils the root causes of reward hacking and overconfidence in current RL paradigms incorporating uncertainty-based rewards, based on which we propose the UnCertainty-Aware Policy Optimization (UCPO) framework. UCPO employs Ternary Advantage Decoupling to separate and independently normalize deterministic and uncertain rollouts, thereby eliminating advantage bias. Furthermore, a Dynamic Uncertainty Reward Adjustment mechanism adapts uncertainty weights in real-time according to model evolution and instance difficulty. Experimental results in mathematical reasoning and general tasks demonstrate that UCPO effectively resolves the reward imbalance, significantly improving the reliability of the model beyond their knowledge boundaries.
title UCPO: Uncertainty-Aware Policy Optimization
topic Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2601.22648