MMBU: A Massive Multi-modal Biomedical Understanding Benchmark to Probe the Perception Capabilities of Vision-Language Models

Name: MMBU: A Massive Multi-modal Biomedical Understanding Benchmark to Probe the Perception Capabilities of Vision-Language Models
Uploaded: 2026-10-06T07:56:20.247245+00:00
Duration: 336 s
Description: MMBU is a massive benchmark testing vision-language models (VLMs) across 35 biomedical submodalities and various imaging modalities. It aims to rigorously evaluate model perception by providing rich metadata and diverse tasks, revealing critical weaknesses in current VLMs, such as poor object detection and low accuracy in classification.

MMBU is a massive benchmark testing vision-language models (VLMs) across 35 biomedical submodalities and various imaging modalities. It aims to rigorously evaluate model perception by providing rich metadata and diverse tasks, revealing critical weaknesses in current VLMs, such as poor object detection and low accuracy in classification.

Open the full episode