Patch-level Visual Referring Prompting (VPR) on 50 manually annotated pathology images (test)
96Win Rate (GPT-4o)CPath-Omni
Evaluation results
| Method | Links | ||
|---|---|---|---|
| CPath-OmniComparison Model=PathGen-LLaVA2024.12 | 96 | 98 | |
| CPath-OmniComparison Model=LLaVA-Med2024.12 | 96 | 100 | |
| CPath-OmniComparison Model=Quilt-LLaVA2024.12 | 84 | 90 |