Image de couverture de livre

Confidence Before Answering: A Paradigm Shift for Efficient LLM Uncertainty Estimation

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Li, Changcheng, Wu, Jiancan, Zhang, Hengheng, Chen, Zhengsu, An, Guo, Qiu, Junxiang, Wang, Xiang, Tian, Qi
Format:	Preprint
Publié:	2026
Sujets:	Computation and Language
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

_version_	1866912947451723776
author	Li, Changcheng Wu, Jiancan Zhang, Hengheng Chen, Zhengsu An, Guo Qiu, Junxiang Wang, Xiang Tian, Qi
author_facet	Li, Changcheng Wu, Jiancan Zhang, Hengheng Chen, Zhengsu An, Guo Qiu, Junxiang Wang, Xiang Tian, Qi
contents	Reliable deployment of large language models (LLMs) requires accurate uncertainty estimation. Existing methods are predominantly answer-first, producing confidence only after generating an answer, which measure the correctness of a specific response and limits practical usability. We study a confidence-first paradigm, where the model outputs its confidence before answering, interpreting this score as the model's probability of answering the question correctly under its current policy. We propose CoCA(Co-optimized Confidence and Answers), a GRPO reinforcement learning framework that jointly optimizes confidence calibration and answer accuracy via segmented credit assignment. By assigning separate rewards and group-relative advantages to confidence and answer segments, CoCA enables stable joint optimization and avoids reward hacking. Experiments across math, code, and factual QA benchmarks show improved calibration and uncertainty discrimination while preserving answer quality, thereby enabling a broader range of downstream applications.
format	Preprint
id	arxiv_https___arxiv_org_abs_2603_05881
institution	arXiv
publishDate	2026
record_format	arxiv
spellingShingle	Confidence Before Answering: A Paradigm Shift for Efficient LLM Uncertainty Estimation Li, Changcheng Wu, Jiancan Zhang, Hengheng Chen, Zhengsu An, Guo Qiu, Junxiang Wang, Xiang Tian, Qi Computation and Language Reliable deployment of large language models (LLMs) requires accurate uncertainty estimation. Existing methods are predominantly answer-first, producing confidence only after generating an answer, which measure the correctness of a specific response and limits practical usability. We study a confidence-first paradigm, where the model outputs its confidence before answering, interpreting this score as the model's probability of answering the question correctly under its current policy. We propose CoCA(Co-optimized Confidence and Answers), a GRPO reinforcement learning framework that jointly optimizes confidence calibration and answer accuracy via segmented credit assignment. By assigning separate rewards and group-relative advantages to confidence and answer segments, CoCA enables stable joint optimization and avoids reward hacking. Experiments across math, code, and factual QA benchmarks show improved calibration and uncertainty discrimination while preserving answer quality, thereby enabling a broader range of downstream applications.
title	Confidence Before Answering: A Paradigm Shift for Efficient LLM Uncertainty Estimation
topic	Computation and Language
url	https://arxiv.org/abs/2603.05881