English

SConU: Selective Conformal Uncertainty in Large Language Models

Computation and Language 2025-07-01 v2 Artificial Intelligence Machine Learning Machine Learning

Abstract

As large language models are increasingly utilized in real-world applications, guarantees of task-specific metrics are essential for their reliable deployment. Previous studies have introduced various criteria of conformal uncertainty grounded in split conformal prediction, which offer user-specified correctness coverage. However, existing frameworks often fail to identify uncertainty data outliers that violate the exchangeability assumption, leading to unbounded miscoverage rates and unactionable prediction sets. In this paper, we propose a novel approach termed Selective Conformal Uncertainty (SConU), which, for the first time, implements significance tests, by developing two conformal p-values that are instrumental in determining whether a given sample deviates from the uncertainty distribution of the calibration set at a specific manageable risk level. Our approach not only facilitates rigorous management of miscoverage rates across both single-domain and interdisciplinary contexts, but also enhances the efficiency of predictions. Furthermore, we comprehensively analyze the components of the conformal procedures, aiming to approximate conditional coverage, particularly in high-stakes question-answering tasks.

Keywords

Cite

@article{arxiv.2504.14154,
  title  = {SConU: Selective Conformal Uncertainty in Large Language Models},
  author = {Zhiyuan Wang and Qingni Wang and Yue Zhang and Tianlong Chen and Xiaofeng Zhu and Xiaoshuang Shi and Kaidi Xu},
  journal= {arXiv preprint arXiv:2504.14154},
  year   = {2025}
}

Comments

Accepted by ACL 2025 Main

R2 v1 2026-06-28T23:04:01.165Z