Tabular Classification on OpenML-CC18 HPO subset v2 (test)
100Banknote PerformanceTabPFN v2
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| TabPFN v2Model Category=Foundation models2026.03 | 100 | 69.4 | 84.4 | 86.6 | 73.7 | 84.9 | 82.7 | 98.7 | 85.1 | |
| KNNModel Category=Classical models2026.03 | 100 | 62.5 | 75.1 | 65 | 65.1 | 79.6 | 85.5 | 78.2 | 76.4 | |
| Nano TabPFNModel Category=Foundation models2026.03 | 99.9 | 69.6 | 85.2 | 86 | 74 | 82.2 | 68.9 | 92.7 | 82.3 | |
| Random ForestModel Category=Classical models2026.03 | 99.5 | 67.4 | 81.9 | 84.3 | 72.7 | 85 | 89.9 | 92.5 | 84.2 | |
| ShapPFN (arch)Model Category=Foundation models, Variant=Architecture only (loss weight 0)2026.03 | 99.3 | 69.8 | 85.1 | 85.6 | 72.8 | 82 | 70.2 | 84.2 | 81.1 | |
| ShapPFN (+loss)Model Category=Foundation models, Variant=Proposed full method2026.03 | 99.1 | 70.3 | 83.5 | 86.3 | 72.4 | 82.7 | 72.2 | 85.2 | 81.5 | |
| Decision TreeModel Category=Classical models2026.03 | 93.8 | 62.4 | 65.6 | 68 | 55.8 | 64.3 | 72.9 | 71.4 | 69.3 |