Hallucination Detection and Evaluation of Large Language Model

TL;DR

This paper introduces HHEM, a lightweight and efficient hallucination detection model for LLMs, improving evaluation speed and accuracy while analyzing model stability and hallucination patterns.

Contribution

We propose HHEM, a novel classification-based framework that reduces evaluation time significantly and enhances hallucination detection accuracy across various LLMs.

Findings

HHEM reduces evaluation time from 8 hours to 10 minutes.

HHEM with non-fabrication checking achieves 82.2% accuracy and 78.9% TPR.

Larger models (7B-9B) tend to hallucinate less, but intermediate models are more unstable.

Abstract

Hallucinations in Large Language Models (LLMs) pose a significant challenge, generating misleading or unverifiable content that undermines trust and reliability. Existing evaluation methods, such as KnowHalu, employ multi-stage verification but suffer from high computational costs. To address this, we integrate the Hughes Hallucination Evaluation Model (HHEM), a lightweight classification-based framework that operates independently of LLM-based judgments, significantly improving efficiency while maintaining high detection accuracy. We conduct a comparative analysis of hallucination detection methods across various LLMs, evaluating True Positive Rate (TPR), True Negative Rate (TNR), and Accuracy on question-answering (QA) and summarization tasks. Our results show that HHEM reduces evaluation time from 8 hours to 10 minutes, while HHEM with non-fabrication checking achieves the highest accuracy \(82.2\%\) and TPR \(78.9\%\). However, HHEM struggles with localized hallucinations in summarization tasks. To address this, we introduce segment-based retrieval, improving detection by verifying smaller text components. Additionally, our cumulative distribution function (CDF) analysis indicates that larger models (7B-9B parameters) generally exhibit fewer hallucinations, while intermediate-sized models show higher instability. These findings highlight the need for structured evaluation frameworks that balance computational efficiency with robust factual validation, enhancing the reliability of LLM-generated content.