Arrhythmia classification on MIT-BIH-SUP
61.5Macro F1Threshold-induced routed model
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Threshold-induced routed modelArchitecture=DeepArrhythmia2026.05 | 61.5 | 95.3 | |
| Rich-specialization modelArchitecture=DeepArrhythmia2026.05 | 61.4 | 95.2 | |
| Simple routing baselineArchitecture=DeepArrhythmia2026.05 | 61.3 | 95.3 | |
| SVMSeeds=Mean over 32026.05 | 57.5 | 93.8 | |
| Minimal-specialization modelArchitecture=DeepArrhythmia2026.05 | 56.2 | 93.8 | |
| Signal + image + peaks + features + morphology2026.05 | 53.6 | 92.6 | |
| Gemini 3.1Decoding Temperature=02026.05 | 52.2 | 84.9 | |
| Signal + image + peaks + features2026.05 | 51.6 | 91.2 | |
| xECGSeeds=Mean over 32026.05 | 51.2 | 90.1 | |
| 2D CNNSeeds=Mean over 32026.05 | 48.6 | 82.9 | |
| 1D ResNetSeeds=Mean over 32026.05 | 46.7 | 81.4 | |
| TCNSeeds=Mean over 32026.05 | 45.1 | 78.3 | |
| Qwen3.5-4BDecoding Temperature=02026.05 | 44.3 | 86.5 | |
| TransformerSeeds=Mean over 32026.05 | 42.7 | 91.3 | |
| STFT CNNSeeds=Mean over 32026.05 | 39.5 | 83.2 | |
| Signal + image + peaks2026.05 | 36.1 | 73.9 | |
| GPT-5.4Decoding Temperature=02026.05 | 30.7 | 72.7 | |
| Gemma-3-27BDecoding Temperature=02026.05 | 27.7 | 81.7 | |
| PULSEDecoding Temperature=02026.05 | 22.6 | 66 | |
| GEMDecoding Temperature=02026.05 | 19.4 | 46.8 | |
| ECG-R1Decoding Temperature=02026.05 | 11.6 | 23.7 |