EEG-Bench: A Benchmark for EEG Foundation Models in Clinical Applications

TL;DR

This paper presents EEG-Bench, a comprehensive benchmarking framework for evaluating EEG foundation models across multiple clinical diagnostic tasks, highlighting the performance and robustness of various models.

Contribution

It introduces a standardized, multi-task benchmark for EEG models, including datasets, evaluation protocols, and a comparison of classical and modern foundation models.

Findings

Foundation models perform well in some tasks but are often outperformed by simpler models under distribution shifts.

The benchmark enables reproducibility and fair comparison across models and datasets.

Minimal preprocessing and standardized evaluation facilitate broader adoption.

Abstract

We introduce a unified benchmarking framework focused on evaluating EEG-based foundation models in clinical applications. The benchmark spans 11 well-defined diagnostic tasks across 14 publicly available EEG datasets, including epilepsy, schizophrenia, Parkinson's disease, OCD, and mild traumatic brain injury. It features minimal preprocessing, standardized evaluation protocols, and enables side-by-side comparisons of classical baselines and modern foundation models. Our results show that while foundation models achieve strong performance in certain settings, simpler models often remain competitive, particularly under clinical distribution shifts. To facilitate reproducibility and adoption, we release all prepared data and code in an accessible and extensible format.