FactTest: Factuality Testing in Large Language Models with Finite-Sample and Distribution-Free Guarantees

TL;DR

FactTest is a new statistical framework that rigorously tests the factuality of large language models with guarantees on error rates, improving reliability in high-stakes applications.

Contribution

It introduces a distribution-free, hypothesis testing approach for LLM factuality assessment with strong Type I and Type II error control, applicable to any model and sample size.

Findings

Effectively detects hallucinations in LLMs.

Enhances model abstention on unknown questions.

Over 40% accuracy improvement in experiments.

Abstract

The propensity of Large Language Models (LLMs) to generate hallucinations and non-factual content undermines their reliability in high-stakes domains, where rigorous control over Type I errors (the conditional probability of incorrectly classifying hallucinations as truthful content) is essential. Despite its importance, formal verification of LLM factuality with such guarantees remains largely unexplored. In this paper, we introduce FactTest, a novel framework that statistically assesses whether a LLM can confidently provide correct answers to given questions with high-probability correctness guarantees. We formulate factuality testing as hypothesis testing problem to enforce an upper bound of Type I errors at user-specified significance levels. Notably, we prove that our framework also ensures strong Type II error control under mild conditions and can be extended to maintain its effectiveness when covariate shifts exist. Our approach is distribution-free and works for any number of human-annotated samples. It is model-agnostic and applies to any black-box or white-box LM. Extensive experiments on question-answering (QA) and multiple-choice benchmarks demonstrate that FactTest effectively detects hallucinations and improves the model's ability to abstain from answering unknown questions, leading to an over 40% accuracy improvement.