Structured Pathology Report Prediction on Structured Pathology Report Dataset (test)
68.89Top-1 AccuracyHiPath
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| HiPathEncoder=UNI2, Trainable Parameters=15.0 M, Text-augmented training=true2026.03 | 68.89 | 88.68 | 74.74 | 97.32 | |
| Flat CrossAttnEncoder=UNI2, Trainable Parameters=15.7 M, Text-augmented training=true2026.03 | 55.98 | 83.83 | 68.92 | 92.18 | |
| Deep MLPEncoder=UNI2, Trainable Parameters=15.9 M2026.03 | 32.49 | 73.86 | 61.76 | 87.46 | |
| ABMILEncoder=UNI2, Trainable Parameters=603 K2026.03 | 32.37 | 73.54 | 60.29 | 87.21 | |
| Linear probeEncoder=UNI2, Trainable Parameters=312 K2026.03 | 30.64 | 71.81 | 58.54 | 87.5 | |
| Text-only (no images)Encoder=Qwen3, Trainable Parameters=02026.03 | 21.52 | 54.27 | — | — | |
| Linear probeEncoder=CONCH, Trainable Parameters=156 K2026.03 | 14.63 | 55.71 | 42.68 | 85.62 | |
| Linear probeEncoder=PLIP, Trainable Parameters=156 K2026.03 | 10.64 | 45.13 | 33.91 | 78.15 |