Durability prediction (Task 2) on CMI-PB (test)
0.791AUROCLogistic regression
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Logistic regressionInput features=Raw features (mean-imputed concatenation), Imputation method=per-feature training-set mean imputation2026.05 | 0.791 | 0.555 | |
| TabMLPInput features=Raw features (mean-imputed concatenation), Imputation method=per-feature training-set mean imputation2026.05 | 0.777 | 0.526 | |
| multi-task contrastive multimodal fusion architectureInput features=TabPFN-v2 embeddings (concatenation), Method Label=Preferred (ours)2026.05 | 0.755 | 0.519 | |
| XGBoostInput features=Raw features (mean-imputed concatenation), Imputation method=per-feature training-set mean imputation2026.05 | 0.645 | 0.382 | |
| Logistic regressionInput features=TabPFN-v2 embeddings (concatenation), Embedding dimensions=6,1442026.05 | 0.609 | 0.336 | |
| TabMLPInput features=TabPFN-v2 embeddings (concatenation), Embedding dimensions=6,1442026.05 | 0.5 | 0.5 | |
| XGBoostInput features=TabPFN-v2 embeddings (concatenation), Embedding dimensions=6,1442026.05 | 0.364 | 0.214 |