Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMs

arXiv:2401.06209v2 Announce Type: replace
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. …

abstract arxiv capabilities clip cs.cv good however image instance language language models large language large language models llms multimodal multimodal llms multimodal models pre-training reasoning research stem training type vision visual