Kernel Language Entropy: Fine-grained Uncertainty Quantification for LLMs from Semantic Similarities

TL;DR

This paper introduces Kernel Language Entropy (KLE), a novel semantic uncertainty estimation method for LLMs that improves trustworthiness by capturing fine-grained semantic dependencies and outperforming previous approaches.

Contribution

KLE is a new entropy-based approach that encodes semantic similarities with kernels and provides more detailed uncertainty estimates for LLM outputs.

Findings

KLE outperforms previous methods in uncertainty quantification across multiple datasets.

Theoretically, KLE generalizes the semantic entropy method.

Empirically, KLE improves detection of incorrect model responses.

Abstract

Uncertainty quantification in Large Language Models (LLMs) is crucial for applications where safety and reliability are important. In particular, uncertainty can be used to improve the trustworthiness of LLMs by detecting factually incorrect model responses, commonly called hallucinations. Critically, one should seek to capture the model's semantic uncertainty, i.e., the uncertainty over the meanings of LLM outputs, rather than uncertainty over lexical or syntactic variations that do not affect answer correctness. To address this problem, we propose Kernel Language Entropy (KLE), a novel method for uncertainty estimation in white- and black-box LLMs. KLE defines positive semidefinite unit trace kernels to encode the semantic similarities of LLM outputs and quantifies uncertainty using the von Neumann entropy. It considers pairwise semantic dependencies between answers (or semantic clusters), providing more fine-grained uncertainty estimates than previous methods based on hard clustering of answers. We theoretically prove that KLE generalizes the previous state-of-the-art method called semantic entropy and empirically demonstrate that it improves uncertainty quantification performance across multiple natural language generation datasets and LLM architectures.