Benchmark RadarPathAgentBench
PathAgentBench evaluates vision-language models on whole-slide pathology images across four capabilities: image-to-text matching, text-to-image retrieval, diagnostic-region localization, and multi-scale reasoning. It includes 1,822 TCGA WSIs and 17,135 diagnostic paths.
- Released
- 2026-07-21
- Readiness
- Paper only
- Primary field
- Health & Life Sciences
Why it matters
Most pathology benchmarks use pre-cropped patches, not whole-slide exploration. PathAgentBench provides a unified framework with annotated paths, revealing a significant gap in evidence acquisition and supporting progress in autonomous WSI diagnosis.
Motivation
Whole-slide image (WSI) diagnosis requires identifying diagnostically relevant regions, examining them across magnifications, and integrating multi-scale evidence.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.