VLM-SubtleBench: How Far Are VLMs from Human-Level Subtle Comparative Reasoning?
This benchmark addresses a critical gap in evaluating VLMs' ability to discern subtle visual differences, which is crucial for real-world applications in domains like industrial anomaly detection and medical imaging.
This paper introduces VLM-SubtleBench, a new benchmark to evaluate vision-language models (VLMs) on subtle comparative reasoning across ten difference types and diverse domains like industrial, aerial, and medical imagery. Their evaluation reveals systematic gaps between VLM and human performance, indicating where VLM reasoning sharply deteriorates.
The ability to distinguish subtle differences between visually similar images is essential for diverse domains such as industrial anomaly detection, medical imaging, and aerial surveillance. While comparative reasoning benchmarks for vision-language models (VLMs) have recently emerged, they primarily focus on images with large, salient differences and fail to capture the nuanced reasoning required for real-world applications. In this work, we introduce VLM-SubtleBench, a benchmark designed to evaluate VLMs on subtle comparative reasoning. Our benchmark covers ten difference types - Attribute, State, Emotion, Temporal, Spatial, Existence, Quantity, Quality, Viewpoint, and Action - and curate paired question-image sets reflecting these fine-grained variations. Unlike prior benchmarks restricted to natural image datasets, our benchmark spans diverse domains, including industrial, aerial, and medical imagery. Through extensive evaluation of both proprietary and open-source VLMs, we reveal systematic gaps between model and human performance across difference types and domains, and provide controlled analyses highlighting where VLMs' reasoning sharply deteriorates. Together, our benchmark and findings establish a foundation for advancing VLMs toward human-level comparative reasoning.