Counterfactual Debating with Preset Stances for Hallucination Elimination of LLMs

TL;DR

This paper introduces CFMAD, a counterfactual debate framework where preset stances guide LLMs to generate justifications, enabling better hallucination elimination through structured multi-agent debate and third-party evaluation.

Contribution

The paper proposes a novel multi-agent debate framework with preset stances to override biases and improve hallucination detection in LLMs.

Findings

CFMAD outperforms existing methods on four datasets across three tasks.

Counterfactual debate improves answer rationality and hallucination detection.

Preset stances effectively override LLM biases during answer inspection.

Abstract

Large Language Models (LLMs) excel in various natural language processing tasks but struggle with hallucination issues. Existing solutions have considered utilizing LLMs' inherent reasoning abilities to alleviate hallucination, such as self-correction and diverse sampling methods. However, these methods often overtrust LLMs' initial answers due to inherent biases. The key to alleviating this issue lies in overriding LLMs' inherent biases for answer inspection. To this end, we propose a CounterFactual Multi-Agent Debate (CFMAD) framework. CFMAD presets the stances of LLMs to override their inherent biases by compelling LLMs to generate justifications for a predetermined answer's correctness. The LLMs with different predetermined stances are engaged with a skeptical critic for counterfactual debate on the rationality of generated justifications. Finally, the debate process is evaluated by a third-party judge to determine the final answer. Extensive experiments on four datasets of three tasks demonstrate the superiority of CFMAD over existing methods.