Interpreting and Mitigating Hallucination in MLLMs through Multi-agent Debate

TL;DR

This paper introduces a multi-agent debate framework with self-reflection to interpret, mitigate, and understand hallucinations in multimodal large language models, emphasizing divergent and slow thinking.

Contribution

It proposes a novel multi-agent debate approach combined with self-reflection to reduce hallucinations and interpret their causes in MLLMs, advancing beyond prior detection methods.

Findings

Effective hallucination mitigation across multiple benchmarks

Enhanced interpretability of hallucination causes

Improved evaluation of MLLMs' creativity capabilities

Abstract

MLLMs often generate outputs that are inconsistent with the visual content, a challenge known as hallucination. Previous methods focus on determining whether a generated output is hallucinated, without identifying which image region leads to the hallucination or interpreting why such hallucinations occur. In this paper, we argue that hallucination in MLLMs is partially due to a lack of slow-thinking and divergent-thinking in these models. To address this, we propose adopting a self-reflection scheme to promote slow-thinking. Furthermore, we consider eliminating hallucination as a complex reasoning task and propose a multi-agent debate approach to encourage divergent-thinking. Consequently, our approach can not only mitigate hallucinations but also interpret why they occur and detail the specifics of hallucination. In addition, we propose to distinguish creativity from hallucination in the context of MLLMs, and illustrate how to evaluate MLLMs' creativity capability. Extensive experiments on various benchmarks demonstrate that our approach exhibits generalized hallucinations-mitigating performance across several MLLMs.