ACVR2A Mutation Prediction on ACVR2A mutation (held-out set)
91Weighted F1 ScoreMOOZY
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MOOZYEvaluation Protocol=Frozen model evaluated with MLP probe2026.03 | 91 | 91 | 90 | |
| MOOZYEvaluation Protocol=Frozen-feature MLP probe2026.03 | 91 | 91 | 90 | |
| CHIEFEvaluation Protocol=Frozen-feature MLP probe2026.03 | 89 | 80 | 80 | |
| MadeleineEvaluation Protocol=Frozen-feature MLP probe2026.03 | 89 | 76 | 81 | |
| TITANEvaluation Protocol=Frozen-feature MLP probe2026.03 | 87 | 79 | 76 | |
| PRISMEvaluation Protocol=Frozen-feature MLP probe2026.03 | 85 | 83 | 81 | |
| UNI v2Evaluation Protocol=Aggregator trained from scratch on frozen features, Aggregator=Mean over five MIL architectures (MeanMIL, ABMIL, CLAM, DSMIL, TransMIL)2026.03 | 84 | 72 | 73 | |
| BackboneEvaluation Protocol=Aggregator trained from scratch on frozen features, Aggregator=Mean over five MIL architectures (MeanMIL, ABMIL, CLAM, DSMIL, TransMIL)2026.03 | 83 | 74 | 73 | |
| CONCH v1.5Evaluation Protocol=Aggregator trained from scratch on frozen features, Aggregator=Mean over five MIL architectures (MeanMIL, ABMIL, CLAM, DSMIL, TransMIL)2026.03 | 82 | 70 | 70 | |
| Phikon v2Evaluation Protocol=Aggregator trained from scratch on frozen features, Aggregator=Mean over five MIL architectures (MeanMIL, ABMIL, CLAM, DSMIL, TransMIL)2026.03 | 81 | 73 | 68 | |
| Giga PathEvaluation Protocol=Frozen-feature MLP probe2026.03 | 80 | 74 | 65 | |
| MUSKEvaluation Protocol=Aggregator trained from scratch on frozen features, Aggregator=Mean over five MIL architectures (MeanMIL, ABMIL, CLAM, DSMIL, TransMIL)2026.03 | 79 | 60 | 65 |