EEG/EMG Foundation
EEG/EMG Foundation Challenge · NeurIPS 2026
Four decoding tracks test models across stimuli, sessions, devices, and people.
Biosignal foundation models claim to transfer across subjects, sessions, and devices with limited per-task training. This challenge tests that claim against strong task-specific baselines.
The four-track design separates two claims usually conflated in the literature: single-task performance and shift robustness. One harness and one reproducibility audit hold evaluation fixed, so a per-track score can be attributed to the model rather than the protocol.
Cross-stimulus
Decodes which image a subject is viewing from a single EEG epoch, evaluated as ranking against held-out candidates. Targets are frozen DINOv2-giant embeddings, so the difficulty is in the EEG side, not in re-learning the image space. The controlled shift is cross-stimulus: test stimuli are unseen during training.
Visual retrieval, DINOv2, RSVP, object viewing
Cross-session
Decodes a cued mental command (motor imagery, mental calculation, word association) from short EEG windows. The controlled shift is cross-session: models train on early sessions and are scored on later sessions of the same subject, with no per-session recalibration allowed.
Motor imagery, mental math, word association, session drift
Cross-device
Predicts the latency in seconds from recording start to the first stable N2 epoch, on consumer-grade wearable EEG. The controlled shift is cross-device: home-wearable channels and montages instead of the clinical polysomnography setups of the seed corpora. Precise timing replaces full staging because the wearable signal is too sparse to support per-epoch hypnogram reconstruction.
Sleep onset, wearable EEG, Muse, home monitoring
Cross-user + stage
Regresses 20 hand-joint angle trajectories from 16-channel wrist surface EMG. The controlled shift combines held-out users and movement stages, testing whether a model survives changes in anatomy, device placement, and hand kinematics.
Hand pose, trajectory regression, wristbands, UmeTrack joints
Evidence, held to one protocol
Each track ships with a brand-new corpus alongside curated seed datasets. EEGDash exposes 700+ additional EEG corpora through the same BIDS-first streaming interface.
Four new datasets. One per cognitive domain.
Each track ships with a brand-new corpus, alongside curated seed datasets. For foundation-model pretraining, EEGDash exposes 700+ additional EEG corpora under the same BIDS-first streaming interface.
New 11-subject evaluation cohort recorded with the same Emotiv hardware and natural-images paradigm as the public Alljoined-1.6M corpus. Released by Alljoined.
20 subjects × 6 sessions, mixed Graz and BrainHero paradigms.
~1,000-subject home-wearable EEG with n2_onset annotations.
A 370-hour benchmark pairing 16-channel wrist sEMG with 20 UmeTrack joint angles across 193 participants and 29 movement stages.
The competition runs Sep 16 - Nov 16, 2026 (AoE). Starter kit, baselines, datasets, and the scoring container are released at the start; public validation and sealed evaluation run across one window.
Five warm-up submissions per day. Fixed Codabench H100/H200 container with a 60-minute inference budget.
Two submissions per day in the sealed final phase. Ground truth stays confidential.
Top-3 are re-run on the sealed split with ±2σ tolerance; final-ranking dates are announced later in 2026.
Winners present at Foundation Models for the Brain and Body, the NeurIPS workshop in Sydney.
$2,500 / top-3 / track, sealed evaluation, and a workshop slot in Sydney.
View awardsCompute, datasets, prizes, and engineering hours from 15 institutions across 6 countries. Yneuro hosts the platform; AWS provides the GPU cloud. Each track has its own sponsor: Alljoined for EEG-to-IMG, Meta FAIR Brain & AI for BCI decoding, Muse for sleep onset, and Meta Reality Labs for EMG-to-Pose.
Meet the team
Four tracks, sealed evaluation, a $30,000 prize pool ($2,500 per top-3 team per track), and a workshop slot in Sydney. Sign up to the mailing list for the start-kit and dataset links, submission access when the window opens Sep 16, and the post-summer announcement of audit, prizes, and workshop dates.
Read the rules before you enter. Submission caps, eligibility, the external-data policy and the reproducibility audit are all set out on the Rules & FAQ page. Entering the competition means you accept them.