Saved in:
Bibliographic Details
Main Authors: Yang, Pengyue, Wen, Jiawen, Jin, Haolin, Huang, Linghan, Chen, Huaming, Chen, Ling
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2602.00977
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911413702754304
author Yang, Pengyue
Wen, Jiawen
Jin, Haolin
Huang, Linghan
Chen, Huaming
Chen, Ling
author_facet Yang, Pengyue
Wen, Jiawen
Jin, Haolin
Huang, Linghan
Chen, Huaming
Chen, Ling
contents Large language models (LLMs) are increasingly deployed in domains where errors carry high social, scientific, or safety costs. Yet standard confidence estimators, such as token likelihood, semantic similarity and multi-sample consistency, remain brittle under distribution shift, domain-specialised text, and compute limits. In this work, we present Structural Confidence, a single-pass, model-agnostic framework that enhances output correctness prediction based on multi-scale structural signals derived from a model's final-layer hidden-state trajectory. By combining spectral, local-variation, and global shape descriptors, our method captures internal stability patterns that are missed by probabilities and sentence embeddings. We conduct extensive, cross-domain evaluation across four heterogeneous benchmarks-FEVER (fact verification), SciFact (scientific claims), WikiBio-hallucination (biographical consistency), and TruthfulQA (truthfulness-oriented QA). Our Structural Confidence framework demonstrates strong performance compared with established baselines in terms of AUROC and AUPR. More importantly, unlike sampling-based consistency methods which require multiple stochastic generations and an auxiliary model, our approach uses a single deterministic forward pass, offering a practical basis for efficient, robust post-hoc confidence estimation in socially impactful, resource-constrained LLM applications.
format Preprint
id arxiv_https___arxiv_org_abs_2602_00977
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Trust in One Round: Confidence Estimation for Large Language Models via Structural Signals
Yang, Pengyue
Wen, Jiawen
Jin, Haolin
Huang, Linghan
Chen, Huaming
Chen, Ling
Computation and Language
Machine Learning
Large language models (LLMs) are increasingly deployed in domains where errors carry high social, scientific, or safety costs. Yet standard confidence estimators, such as token likelihood, semantic similarity and multi-sample consistency, remain brittle under distribution shift, domain-specialised text, and compute limits. In this work, we present Structural Confidence, a single-pass, model-agnostic framework that enhances output correctness prediction based on multi-scale structural signals derived from a model's final-layer hidden-state trajectory. By combining spectral, local-variation, and global shape descriptors, our method captures internal stability patterns that are missed by probabilities and sentence embeddings. We conduct extensive, cross-domain evaluation across four heterogeneous benchmarks-FEVER (fact verification), SciFact (scientific claims), WikiBio-hallucination (biographical consistency), and TruthfulQA (truthfulness-oriented QA). Our Structural Confidence framework demonstrates strong performance compared with established baselines in terms of AUROC and AUPR. More importantly, unlike sampling-based consistency methods which require multiple stochastic generations and an auxiliary model, our approach uses a single deterministic forward pass, offering a practical basis for efficient, robust post-hoc confidence estimation in socially impactful, resource-constrained LLM applications.
title Trust in One Round: Confidence Estimation for Large Language Models via Structural Signals
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2602.00977