ESI: Epistemic Uncertainty Quantification via Semantic-preserving Intervention for Large Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Li, Mingda, Li, Xinyu, Zhang, Weinan, Ma, Longxuan
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918160894001152
author Li, Mingda
Li, Xinyu
Zhang, Weinan
Ma, Longxuan
author_facet Li, Mingda
Li, Xinyu
Zhang, Weinan
Ma, Longxuan
contents Uncertainty Quantification (UQ) is a promising approach to improve model reliability, yet quantifying the uncertainty of Large Language Models (LLMs) is non-trivial. In this work, we establish a connection between the uncertainty of LLMs and their invariance under semantic-preserving intervention from a causal perspective. Building on this foundation, we propose a novel grey-box uncertainty quantification method that measures the variation in model outputs before and after the semantic-preserving intervention. Through theoretical justification, we show that our method provides an effective estimate of epistemic uncertainty. Our extensive experiments, conducted across various LLMs and a variety of question-answering (QA) datasets, demonstrate that our method excels not only in terms of effectiveness but also in computational efficiency.
format Preprint
id arxiv_https___arxiv_org_abs_2510_13103
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ESI: Epistemic Uncertainty Quantification via Semantic-preserving Intervention for Large Language Models
Li, Mingda
Li, Xinyu
Zhang, Weinan
Ma, Longxuan
Computation and Language
Artificial Intelligence
Machine Learning
Uncertainty Quantification (UQ) is a promising approach to improve model reliability, yet quantifying the uncertainty of Large Language Models (LLMs) is non-trivial. In this work, we establish a connection between the uncertainty of LLMs and their invariance under semantic-preserving intervention from a causal perspective. Building on this foundation, we propose a novel grey-box uncertainty quantification method that measures the variation in model outputs before and after the semantic-preserving intervention. Through theoretical justification, we show that our method provides an effective estimate of epistemic uncertainty. Our extensive experiments, conducted across various LLMs and a variety of question-answering (QA) datasets, demonstrate that our method excels not only in terms of effectiveness but also in computational efficiency.
title ESI: Epistemic Uncertainty Quantification via Semantic-preserving Intervention for Large Language Models
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2510.13103