GAPP scoring on PPGL five-fold cross-validation
73.4Histologic Pattern (macro-F1)PPGL-Swarm
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| PPGL-SwarmInput=Slide2026.03 | 73.4 | 61.6 | 67.3 | 8.4 | 73 | 1.27 | 71 | 1.2 | |
| SlideChatInput=Slide2026.03 | 73.3 | 54.1 | 60.2 | 12.8 | 55 | 1.74 | 52 | 1.8 | |
| TITANInput=Slide2026.03 | 71.5 | 60.4 | 65.8 | 9.2 | 71 | 1.31 | 68 | 1.4 | |
| MedDrInput=Patch2026.03 | 57.8 | 33.7 | 54.4 | 14.3 | 48 | 1.92 | 45 | 2.1 | |
| GPT-4oInput=Patch2026.03 | 43.3 | 47.2 | 40.6 | 18.2 | 31 | 2.41 | 28 | 2.8 | |
| Claude-4.5-SonnetInput=Patch2026.03 | 30.3 | 55.3 | 49.2 | 17.8 | 33 | 2.28 | 32 | 2.6 | |
| LLaVA-MedInput=Patch2026.03 | 27.6 | 30.1 | 26.3 | 20.1 | 22 | 2.89 | 20 | 3.2 | |
| GPT-4oInput=Thumbnail2026.03 | 21.6 | 12.2 | 7 | 22.5 | 18 | 3.1 | 15 | 3.5 |