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

TL;DR

This paper introduces a confidence-first approach for large language models, enabling them to estimate their answer correctness before generating responses, which improves calibration and practical usability.

Contribution

The paper proposes CoCA, a novel reinforcement learning framework that jointly optimizes confidence calibration and answer accuracy in LLMs, addressing limitations of answer-first methods.

Findings

Improved confidence calibration across benchmarks

Enhanced uncertainty discrimination capabilities

Maintained high answer quality

Abstract

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.