Benchmark RadarAI BENCHMARK PROFILE
PathView-Bench
PathVU is a benchmark for fine-grained multiscale visual understanding in pathology, built from 23 public datasets with human-supervised labels and spatial annotations, covering region and slide fields of view, with 14 VQA-style tasks and 308k samples.
- Released
- 2026-07-30
- Readiness
- Paper only
- Primary field
- Health & Life Sciences
Why it matters
PathVU aims to evaluate fine-grained visual understanding in pathology MLLMs, potentially offering more detailed assessment than existing benchmarks that focus on final diagnoses.
Motivation
Multimodal large language models (MLLMs) are increasingly used to analyze pathology images.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.