ESI: Epistemic Uncertainty Quantification via Semantic-preserving Intervention for Large Language Models
This work addresses reliability issues in large language models for users in AI and NLP by providing a novel uncertainty quantification method, though it appears incremental as it builds on existing causal perspectives.
The paper tackled uncertainty quantification in large language models by linking uncertainty to invariance under semantic-preserving interventions, proposing a grey-box method that measures output variation, and demonstrated effectiveness and computational efficiency across various models and QA datasets.
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.