Self-Captioning Multimodal Interaction Tuning: Amplifying Exploitable Redundancies for Robust Vision Language Models

TL;DR

This paper proposes a self-captioning approach with a Multimodal Interaction Gate to amplify shared information between modalities, reducing hallucinations and improving robustness in vision-language models.

Contribution

It introduces a novel self-captioning workflow that converts unique modality interactions into redundant ones, enhancing model reliability.

Findings

Redundancy amplification reduces visual hallucinations by 38.3%.

Model consistency improves by 16.8% with increased redundancy.

The Multimodal Interaction Gate effectively converts unique interactions into shared information.

Abstract

Current vision language models face hallucination and robustness issues against ambiguous or corrupted modalities. We hypothesize that these issues can be addressed by exploiting the shared information between modalities to compensate for the impaired one. To this end, we analyze multimodal interactions -- redundant (shared), unique (exclusive), and synergistic (emergent) task-relevant information provided by the modalities -- to determine their impacts on model reliability. Specifically, amplifying redundant interactions would increase this exploitable shared information to resolve these issues; yet, modern instruction datasets often eliminate redundancies to prioritize visual grounding. We bridge this gap through a self-captioning workflow featuring a \textsc{Multimodal Interaction Gate}: a mechanism to convert unique interactions into redundant interactions. Our findings suggest that increasing redundancy can reduce visual induced errors by 38.3\% and improve consistency by 16.8\%.