Sleep Stages Classification on HMC
74.79Balanced AccuracySpecMoE
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SpecMoEModel Category=Foundation Models, # Params=4.3 M2026.03 | 74.79 | 75.03 | — | |
| CSBrainModel Category=Foundation Models, # Params=24.3 M2026.03 | 73.45 | 75.06 | — | |
| LaBraMModel Category=Foundation Models, # Params=5.8 M2026.03 | 72.77 | 74.54 | — | |
| CBramodModel Category=Foundation Models, # Params=20.3 M2026.03 | 72.69 | 73.95 | — | |
| EEGConfModel Category=Task-Specific Models2026.03 | 71.49 | 70.8 | — | |
| EEGNetModel Category=Task-Specific Models2026.03 | 65.34 | 65.36 | — | |
| FFCLModel Category=Task-Specific Models2026.03 | 44.27 | 29.02 | — | |
| PanLUNAAdaptation Strategy=Full fine-tuning (FF), Training Modality=EEG+ECG+PPG, Test Modality=EEG2026.04 | 0.7416 | 0.7659 | 0.6946 | |
| KAST-BAR-LargeModel Parameter=2.2B, General Model=Yes, Multi-Task=Yes2026.05 | 0.739 | 0.762 | 0.705 | |
| CSBrainModel Parameter=4.9M, General Model=Yes, Multi-Task=No2026.05 | 0.735 | 0.751 | 0.682 | |
| PanLUNAQuantization Strategy=Quantization-Aware Training (QAT), Weights=INT8, Activations=INT8, Training Modality=EEG+ECG+PPG, Test Modality=EEG+ECG2026.04 | 0.7347 | 0.7273 | 0.6913 | |
| LaBraM-BaseModel Parameter=5.8M, General Model=Yes, Multi-Task=No2026.05 | 0.729 | 0.755 | 0.681 | |
| PhysioOmniTraining Modality=EEG+EOG+EMG, Test Modality=EEG2026.04 | 0.7289 | 0.7635 | 0.688 | |
| LaBraM-BaseTraining Modality=EEG, Test Modality=EEG2026.04 | 0.7286 | 0.7554 | 0.6812 | |
| CBraModModel Parameter=4.0M, General Model=Yes, Multi-Task=No2026.05 | 0.727 | 0.74 | 0.669 | |
| KAST-BAR-BaseModel Parameter=0.8B, General Model=Yes, Multi-Task=Yes2026.05 | 0.725 | 0.749 | 0.691 | |
| CBraModTraining Modality=EEG, Test Modality=EEG2026.04 | 0.7177 | 0.7388 | 0.6653 | |
| PanLUNAPrecision=FP32, Training Modality=EEG+ECG+PPG, Test Modality=EEG+ECG2026.04 | 0.717 | 0.7499 | 0.6699 | |
| EEGPTModel Parameter=25M, General Model=Yes, Multi-Task=Yes2026.05 | 0.703 | 0.732 | 0.658 | |
| PanLUNAAdaptation Strategy=Full fine-tuning (FF), Training Modality=EEG+ECG+PPG, Test Modality=EEG+ECG2026.04 | 0.7002 | 0.7383 | 0.6561 | |
| BIOTTraining Modality=EEG, Test Modality=EEG2026.04 | 0.6862 | 0.7091 | 0.6295 | |
| BIOTModel Parameter=3.2M, General Model=Yes, Multi-Task=No2026.05 | 0.686 | 0.709 | 0.63 | |
| THD-BAR-HugeModel Parameter=1.6B, General Model=Yes, Multi-Task=Yes2026.05 | 0.684 | 0.718 | 0.635 | |
| EEG-ConformerTraining Modality=EEG, Test Modality=EEG2026.04 | 0.6767 | 0.655 | 0.5886 | |
| EEGNetModel Parameter=-, General Model=No, Multi-Task=No2026.05 | 0.622 | 0.628 | 0.561 | |
| ST-TransformerModel Parameter=3.5M, General Model=No, Multi-Task=No2026.05 | 0.595 | 0.615 | 0.5 | |
| NeuroLM-XLModel Parameter=1.7B, General Model=Yes, Multi-Task=Yes2026.05 | 0.576 | 0.588 | 0.48 | |
| SPaRCNetModel Parameter=0.79M, General Model=No, Multi-Task=No2026.05 | 0.554 | 0.57 | 0.479 | |
| PanLUNAQuantization Strategy=Post-Training Quantization (PTQ), Weights=INT8, Activations=INT8, Training Modality=EEG+ECG+PPG, Test Modality=EEG+ECG2026.04 | 0.4221 | 0.4479 | 0.3162 | |
| PanLUNAAdaptation Strategy=Full fine-tuning (FF), Training Modality=EEG+ECG+PPG, Test Modality=ECG2026.04 | 0.2977 | 0.2876 | 0.1095 |