Confidence Estimation for LLMs in Multi-turn Interactions

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Caiqi, Yang, Ruihan, Zhu, Xiaochen, Li, Chengzu, Hu, Tiancheng, Dong, Yijiang River, Yang, Deqing, Collier, Nigel
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911682709684224
author Zhang, Caiqi
Yang, Ruihan
Zhu, Xiaochen
Li, Chengzu
Hu, Tiancheng
Dong, Yijiang River
Yang, Deqing
Collier, Nigel
author_facet Zhang, Caiqi
Yang, Ruihan
Zhu, Xiaochen
Li, Chengzu
Hu, Tiancheng
Dong, Yijiang River
Yang, Deqing
Collier, Nigel
contents While confidence estimation is a promising direction for mitigating hallucinations in Large Language Models (LLMs), current research overwhelmingly focuses on single-turn settings. The dynamics of model confidence in multi-turn conversations, where context accumulates and ambiguity is progressively resolved, remain largely unexplored. This work presents the first systematic study of confidence estimation in multi-turn interactions, establishing a formal evaluation framework grounded in two key desiderata: per-turn calibration and monotonicity of confidence as more information becomes available. To facilitate this, we introduce novel metrics, including a length-normalized Expected Calibration Error (InfoECE), and a new "Hinter-Guesser" paradigm for generating controlled evaluation datasets. Our experiments reveal that widely-used confidence techniques struggle with calibration and monotonicity in multi-turn dialogues. In contrast, a novel logit-based probe we introduce, P(Sufficient), proves comparatively more effective, robustly tracking evidence accumulation and distinguishing it from conversational filler. Our work provides a foundational methodology for developing more reliable and trustworthy conversational agents.
format Preprint
id arxiv_https___arxiv_org_abs_2601_02179
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Confidence Estimation for LLMs in Multi-turn Interactions
Zhang, Caiqi
Yang, Ruihan
Zhu, Xiaochen
Li, Chengzu
Hu, Tiancheng
Dong, Yijiang River
Yang, Deqing
Collier, Nigel
Computation and Language
While confidence estimation is a promising direction for mitigating hallucinations in Large Language Models (LLMs), current research overwhelmingly focuses on single-turn settings. The dynamics of model confidence in multi-turn conversations, where context accumulates and ambiguity is progressively resolved, remain largely unexplored. This work presents the first systematic study of confidence estimation in multi-turn interactions, establishing a formal evaluation framework grounded in two key desiderata: per-turn calibration and monotonicity of confidence as more information becomes available. To facilitate this, we introduce novel metrics, including a length-normalized Expected Calibration Error (InfoECE), and a new "Hinter-Guesser" paradigm for generating controlled evaluation datasets. Our experiments reveal that widely-used confidence techniques struggle with calibration and monotonicity in multi-turn dialogues. In contrast, a novel logit-based probe we introduce, P(Sufficient), proves comparatively more effective, robustly tracking evidence accumulation and distinguishing it from conversational filler. Our work provides a foundational methodology for developing more reliable and trustworthy conversational agents.
title Confidence Estimation for LLMs in Multi-turn Interactions
topic Computation and Language
url https://arxiv.org/abs/2601.02179