Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMs

TL;DR

This paper investigates the visual shortcomings of multimodal large language models, introduces a benchmark to evaluate their visual reasoning, and proposes a method to improve their visual grounding by integrating self-supervised visual features.

Contribution

It identifies systematic visual shortcomings in current multimodal LLMs, creates the MMVP benchmark, and proposes a Mixture of Features approach to enhance visual grounding.

Findings

State-of-the-art models struggle with basic visual patterns.

MMVP benchmark reveals specific visual reasoning errors.

Integrating self-supervised visual features improves visual grounding.

Abstract

Is vision good enough for language? Recent advancements in multimodal models primarily stem from the powerful reasoning abilities of large language models (LLMs). However, the visual component typically depends only on the instance-level contrastive language-image pre-training (CLIP). Our research reveals that the visual capabilities in recent multimodal LLMs (MLLMs) still exhibit systematic shortcomings. To understand the roots of these errors, we explore the gap between the visual embedding space of CLIP and vision-only self-supervised learning. We identify ''CLIP-blind pairs'' - images that CLIP perceives as similar despite their clear visual differences. With these pairs, we construct the Multimodal Visual Patterns (MMVP) benchmark. MMVP exposes areas where state-of-the-art systems, including GPT-4V, struggle with straightforward questions across nine basic visual patterns, often providing incorrect answers and hallucinated explanations. We further evaluate various CLIP-based vision-and-language models and found a notable correlation between visual patterns that challenge CLIP models and those problematic for multimodal LLMs. As an initial effort to address these issues, we propose a Mixture of Features (MoF) approach, demonstrating that integrating vision self-supervised learning features with MLLMs can significantly enhance their visual grounding capabilities. Together, our research suggests visual representation learning remains an open challenge, and accurate visual grounding is crucial for future successful multimodal systems.