Enhancing Visual-Language Modality Alignment in Large Vision Language Models via Self-Improvement

TL;DR

This paper introduces SIMA, a self-improvement framework for large vision-language models that enhances modality alignment using self-generated responses and internal critic mechanisms, without external data dependencies.

Contribution

SIMA is a novel self-improvement approach that improves LVLMs' modality alignment by leveraging existing datasets and internal critics, avoiding external models or data.

Findings

Significant performance improvements on 14 benchmarks.

Outperforms previous methods in modality alignment.

Effective self-critic with new visual metrics.

Abstract

Large vision-language models (LVLMs) have achieved impressive results in visual question-answering and reasoning tasks through vision instruction tuning on specific datasets. However, there remains significant room for improvement in aligning visual and language modalities. Existing methods often depend on external models or data, leading to uncontrollable and unstable alignment results. In this paper, we propose SIMA, a self-improvement framework that enhances visual and language modality alignment without external dependencies. SIMA leverages existing vision instruction tuning datasets to self-generate responses, incorporating an in-context self-critic mechanism that constructs preference pairs for tuning. Crucially, our approach allows LVLMs to act as critics by designing effective critic prompts, eliminating the need for additional fine-tuning with external instruction data. We introduce three novel visual metrics within the self-critic process to guide judgment, significantly improving the accuracy of self-critic. Through extensive experiments across 14 hallucination and comprehensive benchmarks, we demonstrate that SIMA significantly improves LVLM's performance and outperforms previous approaches, achieving superior modality alignment.