How to Detect Hallucinations in LLMs

Source: https://towardsdatascience.com/real-time-llm-hallucination-detection-9a68bb292698 Published: 31 December 2023 · Author: Iulia Brezeanu

Summary

A hands-on benchmark of four self-consistency methods for detecting hallucination without a reference answer, compared on both accuracy and latency.

The Four Methods

MethodModel usedLatency
Sentence-embedding cosine distanceall-MiniLM-L6-v2~0.002 s
SelfCheckGPT-BERTScoreRoBERTa-large (17 layers), MNLI~2 s
SelfCheckGPT-NLIDeBERTa-v3-large, MNLI~1 s
SelfCheckGPT-Promptgpt-3.5-turbo~0.5 s — best performing

BERTScore uses a baseline tensor from 1M Common Crawl sentence pairs; NLI classifies entailment / contradiction / neutral and takes the contradiction probability.

Separation Achieved

On a fabricated subject vs a real one: cosine distance 0.52 vs 0.93; SelfCheckGPT-Prompt 0.0 vs 0.95.

Evaluation Data

WikiBio — 238 Wikipedia topics; of 1,908 annotated sentences, ~40% major-inaccurate, 33% minor-inaccurate, 27% accurate. Inter-annotator agreement (Cohen’s κ) 0.595.