For leaderboard submissions, please contact the authors.

Leaderboard

Official Omni-iEEG benchmark baselines from the ICLR 2026 paper.

Official baseline results for Omni-iEEG. Task 1 is ranked by macro-F1, Task 2 is ranked by pathological channel AUC, and exploratory tasks are ranked by macro-F1. Higher is better for all metrics.

Event Classification F1 0.8061 PyHFO-Omni
Channel AUC 0.8061 TimeConv-CNN
Outcome AUC 0.7438 PyHFO-OmnispkHFO
Ictal F1 0.9245 CLAP

Task 1: Pathological Event Classification

Ranked by macro-F1 over artifact, non-spkHFO, and spkHFO event classes.

Rank Model Precision Recall F1 AUC
1 PyHFO-Omni 0.8025 0.8110 0.8061 0.9390
2 PatchTST Transformer 0.7757 0.7686 0.7726 0.9311
3 TimesNet 0.7589 0.7726 0.7652 0.9221
4 LSTM+Attention 0.7352 0.7359 0.7338 0.9109

Source: Table 4 in the Omni-iEEG paper.

Task 2: Pathological Brain Region Identification

Ranked by pathological channel classification AUC; outcome AUC evaluates resection-ratio outcome prediction.

Rank Model Type Precision Recall F1 Specificity Channel AUC Outcome AUC
1 TimeConv-CNN Segment-based 0.6259 0.7454 0.6469 0.8230 0.8061 0.7380
2 SEEG-NET Segment-based 0.5790 0.7169 0.5259 0.6049 0.7850 0.5952
3 CLAP Segment-based 0.5936 0.6997 0.6009 0.7823 0.7684 0.6770
4 PyHFO-OmnispkHFO Event-based 0.5799 0.6991 0.5635 0.6951 0.7351 0.7438
5 eHFO Event-based 0.6053 0.6466 0.6195 0.4101 0.6611 0.4521
6 PyHFOspkHFO Event-based 0.6000 0.6431 0.6140 0.4089 0.6557 0.4972

Source: Table 5 in the Omni-iEEG paper.

Exploratory Tasks

Ranked by macro-F1. Acc denotes balanced accuracy.

Anatomical Location (5-class)

Rank Model F1 Acc
1 TimeConv-CNN 0.4788 0.4708
2 CLAP 0.4750 0.4894
3 SEEG-NET 0.2520 0.2550

Anatomical Location (12-class)

Rank Model F1 Acc
1 CLAP 0.3540 0.3897
2 TimeConv-CNN 0.3087 0.3215
3 SEEG-NET 0.1081 0.1853

Ictal Period

Rank Model F1 Acc
1 CLAP 0.9245 0.9323
2 TimeConv-CNN 0.8533 0.8720
3 SEEG-NET 0.7526 0.7624

Sleep-Awake

Rank Model F1 Acc
1 CLAP 0.7225 0.9331
2 TimeConv-CNN 0.7118 0.9291
3 SEEG-NET 0.6773 0.6421

Source: Table 8 in the Omni-iEEG paper.