Vision language models are blind: Failing to translate detailed visual features into words

TL;DR

Large vision-language models excel at high-level tasks but struggle with low-level, spatially precise visual tasks, revealing a gap in translating detailed visual features into words.

Contribution

This paper demonstrates that current VLMs fail on simple spatial tasks due to inability to decode visual details into language, highlighting a critical limitation.

Findings

VLMs achieve only 58.07% accuracy on BlindTest tasks.

Performance improves to 77.84% with Claude 3.5 Sonnet.

VLMs struggle with precise spatial information but succeed when shapes are separated.

Abstract

While large language models with vision capabilities (VLMs), e.g., GPT-4o and Gemini 1.5 Pro, score high on many vision-understanding benchmarks, they are still struggling with low-level vision tasks that are easy to humans. Specifically, on BlindTest, our suite of 7 very simple tasks, including identifying (a) whether two circles overlap; (b) how many times two lines intersect; (c) which letter is being circled in a word; and (d) the number of circles in an Olympic-like logo, four state-of-the-art VLMs are only 58.07% accurate on average. Claude 3.5 Sonnet performs the best at 77.84% accuracy, far from the human expected accuracy of 100%. Across different image resolutions and line widths, VLMs including slow-thinking models consistently struggle with those tasks that require precise spatial information when geometric primitives overlap or are close. Yet, VLMs perform at near-100% accuracy when much more space is added to separate shapes and letters. Linear probing experiments show that vision encoders contain sufficient visual information to solve BlindTest and that language models fail to decode this information into correct answers. Code and data are at: https://vlmsareblind.github.io