{\mu}-Bench: A Vision-Language Benchmark for Microscopy Understanding

TL;DR

{b0}-Bench is a comprehensive, expert-curated benchmark for evaluating vision-language models in microscopy, revealing current limitations and proposing solutions to improve biomedical image understanding.

Contribution

The paper introduces {b0}-Bench, the first large-scale, diverse microscopy vision-language benchmark, and evaluates models, highlighting challenges and potential strategies like weight interpolation.

Findings

Current models perform poorly across microscopy tasks.

Specialist models often underperform compared to generalist models.

Fine-tuning can cause catastrophic forgetting of prior knowledge.

Abstract

Recent advances in microscopy have enabled the rapid generation of terabytes of image data in cell biology and biomedical research. Vision-language models (VLMs) offer a promising solution for large-scale biological image analysis, enhancing researchers' efficiency, identifying new image biomarkers, and accelerating hypothesis generation and scientific discovery. However, there is a lack of standardized, diverse, and large-scale vision-language benchmarks to evaluate VLMs' perception and cognition capabilities in biological image understanding. To address this gap, we introduce {\mu}-Bench, an expert-curated benchmark encompassing 22 biomedical tasks across various scientific disciplines (biology, pathology), microscopy modalities (electron, fluorescence, light), scales (subcellular, cellular, tissue), and organisms in both normal and abnormal states. We evaluate state-of-the-art biomedical, pathology, and general VLMs on {\mu}-Bench and find that: i) current models struggle on all categories, even for basic tasks such as distinguishing microscopy modalities; ii) current specialist models fine-tuned on biomedical data often perform worse than generalist models; iii) fine-tuning in specific microscopy domains can cause catastrophic forgetting, eroding prior biomedical knowledge encoded in their base model. iv) weight interpolation between fine-tuned and pre-trained models offers one solution to forgetting and improves general performance across biomedical tasks. We release {\mu}-Bench under a permissive license to accelerate the research and development of microscopy foundation models.