Semantic Entropy Probes: Robust and Cheap Hallucination Detection in LLMs

Jiatong Han, Jannik Kossen, Muhammed Razzak, Lisa Schut, Shreshth A Malik, Yarin Gal

Keywords: uncertainty estimation; large language models; hallucinations; linear probing; interpretability

Abstract: We propose semantic entropy probes (SEPs), a cheap and reliable method for uncertainty quantification in Large Language Models (LLMs). Hallucinations, which are plausible-sounding but factually incorrect and arbitrary model generations, present a major challenge to the practical adoption of LLMs. Recent work by Kuhn et al. proposes semantic entropy (SE), which can reliably detect hallucinations by quantifying the uncertainty over different generations by estimating entropy over semantically equivalent sets of outputs. However, the 5-to-10-fold increase in computation cost associated with SE computation hinders practical adoption. To address this, we propose SEPs, which directly approximate SE from the hidden states of a single generation. SEPs are simple to train and do not require sampling multiple model generations at test time, reducing the overhead of semantic uncertainty quantification to almost zero. We show that SEPs retain high performance for hallucination detection and generalize better to out-of-distribution data than previous probing methods that directly predict model accuracy. Our results across models and tasks suggest that model hidden states capture SE, and our ablation studies give further insights into the token positions and model layers for which this is the case.

Submission Number: 79