Detecting Hallucination and Coverage Errors in Retrieval Augmented Generation for Controversial Topics

TL;DR

This paper presents methods to detect hallucination and coverage errors in retrieval-augmented generation for controversial topics, demonstrating high accuracy with LLM-based classifiers trained on synthetic errors.

Contribution

It introduces and evaluates three error detection methods, notably LLM-based classifiers, for improving factual accuracy in retrieval-augmented LLM responses on controversial topics.

Findings

LLM-based classifiers achieve ROC AUC scores of 95.3% for hallucination detection.

Detection methods perform well even without training data, with over 84% accuracy.

The approach helps identify and mitigate errors in LLM-generated content based on retrieved perspectives.

Abstract

We explore a strategy to handle controversial topics in LLM-based chatbots based on Wikipedia's Neutral Point of View (NPOV) principle: acknowledge the absence of a single true answer and surface multiple perspectives. We frame this as retrieval augmented generation, where perspectives are retrieved from a knowledge base and the LLM is tasked with generating a fluent and faithful response from the given perspectives. As a starting point, we use a deterministic retrieval system and then focus on common LLM failure modes that arise during this approach to text generation, namely hallucination and coverage errors. We propose and evaluate three methods to detect such errors based on (1) word-overlap, (2) salience, and (3) LLM-based classifiers. Our results demonstrate that LLM-based classifiers, even when trained only on synthetic errors, achieve high error detection performance, with ROC AUC scores of 95.3% for hallucination and 90.5% for coverage error detection on unambiguous error cases. We show that when no training data is available, our other methods still yield good results on hallucination (84.0%) and coverage error (85.2%) detection.