Seizure detection on CHB-MIT
82.66Balanced AccuracyCAMEL-CLIP
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CAMEL-CLIPEvaluation Protocol=Trained Classifier Head2026.02 | 82.66 | 38.12 | — | |
| CAMEL-CLIPEvaluation Protocol=Linear Probing2026.02 | 79.88 | 36.89 | — | |
| LaBraM-Base2026.02 | 70.75 | 32.87 | — | |
| BIOT2026.02 | 70.68 | 32.77 | — | |
| EEG-CLIPEvaluation Protocol=Trained Classifier Head2026.02 | 68.43 | 21.91 | — | |
| FFCL2026.02 | 62.62 | 20.49 | — | |
| EEGNet2026.02 | 56.58 | 19.14 | — | |
| EEG-CLIPEvaluation Protocol=Linear Probing2026.02 | 50 | 9.81 | — | |
| MTEEG-DC# Trainable Params=1.1M, Training Paradigm=Multi-task2026.04 | 0.8657 | 0.9831 | 0.9763 | |
| MTEEG-SP# Trainable Params=1.8M, Training Paradigm=Multi-task2026.04 | 0.8586 | 0.9742 | 0.9656 | |
| BIOT# Trainable Params=3.2M, Training Paradigm=Single-task2026.04 | 0.8439 | 0.9367 | 0.9026 | |
| SPaRCNet# Trainable Params=0.79M, Training Paradigm=Single-task2026.04 | 0.8417 | 0.9364 | 0.9151 | |
| LaBraM# Trainable Params=0.3M, Training Paradigm=Single-task, Finetuning Strategy=LoRA2026.04 | 0.8244 | 0.9329 | 0.9105 | |
| ST-Transformer# Trainable Params=3.5M, Training Paradigm=Single-task2026.04 | 0.8229 | 0.9165 | 0.8942 | |
| LaBraM# Trainable Params=5.8M, Training Paradigm=Single-task, Finetuning Strategy=Full2026.04 | 0.8229 | 0.926 | 0.8989 | |
| FFCL# Trainable Params=2.4M, Training Paradigm=Single-task2026.04 | 0.8106 | 0.9225 | 0.8918 | |
| ContraWR# Trainable Params=1.6M, Training Paradigm=Single-task2026.04 | 0.8034 | 0.9057 | 0.8671 | |
| CNN-Transformer# Trainable Params=3.2M, Training Paradigm=Single-task2026.04 | 0.7861 | 0.9032 | 0.8701 | |
| MTEEG-RT# Trainable Params=1.8M, Training Paradigm=Multi-task2026.04 | 0.7637 | 0.9054 | 0.8925 | |
| HPS# Trainable Params=6.0M, Training Paradigm=Multi-task2026.04 | 0.7524 | 0.9223 | 0.8914 | |
| BIOTModel Size=3.2M, Training Setting=With Multiple Dataset Pretraining2025.02 | 0.7068 | 0.3277 | 0.8761 | |
| TFM-TokenizerModel Size=1.9M, Training Setting=Single Dataset Setting2025.02 | 0.675 | 0.3379 | 0.8839 | |
| CBraMod†Model Size=4M, Training Setting=With Multiple Dataset Pretraining2025.02 | 0.6646 | 0.3469 | 0.9071 | |
| EEGPTModel Size=4.7M, Training Setting=With Multiple Dataset Pretraining2025.02 | 0.6644 | 0.3373 | 0.8185 | |
| Vanilla BIOTModel Size=3.2M, Training Setting=Single Dataset Setting2025.02 | 0.664 | 0.2573 | 0.8646 | |
| BIOT*Model Size=3.2M, Training Setting=Single Dataset Setting2025.02 | 0.6582 | 0.3127 | 0.8456 | |
| TFM-Tokenizer †Model Size=1.9M, Training Setting=With Multiple Dataset Pretraining2025.02 | 0.6471 | 0.3554 | 0.8818 | |
| CNN-TransformerModel Size=3.2M, Training Setting=Single Dataset Setting2025.02 | 0.6389 | 0.2479 | 0.8662 | |
| ContraWRModel Size=1.6M, Training Setting=Single Dataset Setting2025.02 | 0.6344 | 0.2264 | 0.8097 | |
| FFCLModel Size=2.4M, Training Setting=Single Dataset Setting2025.02 | 0.6262 | 0.2049 | 0.8271 | |
| ST-TransformerModel Size=3.5M, Training Setting=Single Dataset Setting2025.02 | 0.5915 | 0.1422 | 0.8237 | |
| SPaRCNetModel Size=0.79M, Training Setting=Single Dataset Setting2025.02 | 0.5876 | 0.1247 | 0.8143 | |
| LaBraM-Base†Model Size=5.8M, Training Setting=With Multiple Dataset Pretraining2025.02 | 0.526 | 0.2138 | 0.775 | |
| LaBraM-Base*Model Size=5.8M, Training Setting=Single Dataset Setting2025.02 | 0.5035 | 0.0959 | 0.6624 |