aiqbench analysis
This study rigorously benchmarks six EEG foundation models against classical features and robust negative controls, revealing significant vulnerabilities in dataset identity leakage and performance degradation under population shifts. While one model showed a specific advantage for ictal detection, the results underscore that claims of clinical utility often depend heavily on evaluation methodology rather than genuine generalization capabilities. These findings are critical for researchers assessing the reliability of pretrained models before deploying them in sensitive healthcare settings.
Stress-Testing EEG Foundation Models for Clinical Decoding: Dataset Identity and Targeted Negative Controls
arXiv:2607.24519v1 Announce Type: cross Abstract: Pretrained EEG foundation models are increasingly proposed for clinical decoding, but their transfer across populations and robustness to negative controls remain unclear. We benchmark six models (LaBraM, EEGMamba, CBraMod, REVE, BENDR, and BIOT) on five clinical tasks across four datasets using frozen linear probes with leave-one-subject-out, subject-grouped, or explicitly identified recording-level splits. Selected REVE findings are tested against random initialisation, random features, label permutation, scrambled-label fine-tuning, and projection sensitivity. On Korean dementia (CAUEEG, three-way), frozen REVE reaches 0.568 AUROC versus 0.769 for classical features; the ordering persists on a patient-disjoint held-out split (0.565 versus 0.768). Dataset identity is readily decoded from frozen embeddings (AUROC 1.000 at PCA-50; 0.9998 after band restriction and per-epoch z-scoring), whereas the same PCA-50 pipeline decodes Korean diagnosis at 0.528. A randomly initialised encoder also outperforms pretrained REVE on this task (0.659 versus 0.570). On Alzheimer's disease, Gaussian random projection and PCA of the same pretrained embeddings perform similarly, and classical features nominally exceed REVE at the subject level. The clearest controlled positive is cross-subject ictal detection on CHB-MIT (n=23), where REVE achieves 0.793 AUROC, 9.2 percentage points above a randomly initialised encoder. These results show that EEG foundation-model conclusions depend strongly on evaluation unit, dataset shift, comparator strength, and targeted controls.
Abstract:Pretrained EEG foundation models are increasingly proposed for clinical decoding, but their transfer across populations and robustness to negative controls remain unclear. We benchmark six models (LaBraM, EEGMamba, CBraMod, REVE, BENDR, and BIOT) on five clinical tasks across four datasets using frozen linear probes with leave-one-subject-out, subject-grouped, or explicitly identified recording-level splits. Selected REVE findings are tested against random initialisation, random features, label permutation, scrambled-label fine-tuning, and projection sensitivity. On Korean dementia (CAUEEG, three-way), frozen REVE reaches 0.568 AUROC versus 0.769 for classical features; the ordering persists on a patient-disjoint held-out split (0.565 versus 0.768). Dataset identity is readily decoded from frozen embeddings (AUROC 1.000 at PCA-50; 0.9998 after band restriction and per-epoch z-scoring), whereas the same PCA-50 pipeline decodes Korean diagnosis at 0.528. A randomly initialised encoder also outperforms pretrained REVE on this task (0.659 versus 0.570). On Alzheimer's disease, Gaussian random projection and PCA of the same pretrained embeddings perform similarly, and classical features nominally exceed REVE at the subject level. The clearest controlled positive is cross-subject ictal detection on CHB-MIT (n=23), where REVE achieves 0.793 AUROC, 9.2 percentage points above a randomly initialised encoder. These results show that EEG foundation-model conclusions depend strongly on evaluation unit, dataset shift, comparator strength, and targeted controls.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE)
arXiv:2607.24519 [cs.LG]
(or arXiv:2607.24519v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2607.24519 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 308 words
