MIHBench: Benchmarking and Mitigating Multi-Image Hallucinations in Multimodal Large Language Models

TL;DR

This paper introduces MIHBench, a benchmark for evaluating multi-image hallucinations in multimodal large language models, and proposes a Dynamic Attention Balancing method to mitigate these hallucinations, improving model reliability.

Contribution

It is the first systematic study of multi-image hallucinations in MLLMs and presents a novel benchmark along with an attention-based mitigation technique.

Findings

Multi-image hallucinations increase with more input images.

Single-image hallucination tendencies correlate with multi-image hallucinations.

The proposed method reduces hallucination occurrences and improves reasoning stability.

Abstract

Despite growing interest in hallucination in Multimodal Large Language Models, existing studies primarily focus on single-image settings, leaving hallucination in multi-image scenarios largely unexplored. To address this gap, we conduct the first systematic study of hallucinations in multi-image MLLMs and propose MIHBench, a benchmark specifically tailored for evaluating object-related hallucinations across multiple images. MIHBench comprises three core tasks: Multi-Image Object Existence Hallucination, Multi-Image Object Count Hallucination, and Object Identity Consistency Hallucination, targeting semantic understanding across object existence, quantity reasoning, and cross-view identity consistency. Through extensive evaluation, we identify key factors associated with the occurrence of multi-image hallucinations, including: a progressive relationship between the number of image inputs and the likelihood of hallucination occurrences; a strong correlation between single-image hallucination tendencies and those observed in multi-image contexts; and the influence of same-object image ratios and the positional placement of negative samples within image sequences on the occurrence of object identity consistency hallucination. To address these challenges, we propose a Dynamic Attention Balancing mechanism that adjusts inter-image attention distributions while preserving the overall visual attention proportion. Experiments across multiple state-of-the-art MLLMs demonstrate that our method effectively reduces hallucination occurrences and enhances semantic integration and reasoning stability in multi-image scenarios.