NeuroAtlas: Largest EEG Benchmark for Foundation Models
Researchers have introduced NeuroAtlas, the largest EEG benchmark to date, comprising 42 datasets and 260,000 hours of data. It covers clinical EEG applications (epilepsy, sleep medicine, brain age estimation) and brain-computer interfaces, with multiple datasets per task and bespoke clinical evaluation metrics. The benchmark evaluates EEG-specific foundation models (FMs) and generic time-series FMs. Preliminary findings indicate that EEG-specific FMs do not consistently outperform time-series FMs. The study is published on arXiv under identifier 2605.14698.
Key facts
- NeuroAtlas is the largest EEG benchmark to date.
- It includes 42 datasets and 260,000 hours of data.
- Covers clinical EEG (epilepsy, sleep medicine, brain age estimation) and brain-computer interfaces.
- Includes multiple datasets per task and bespoke clinical evaluation metrics.
- Evaluates EEG-specific foundation models and generic time-series foundation models.
- EEG-specific FMs do not consistently outperform time-series FMs.
- Published on arXiv with identifier 2605.14698.
Entities
Institutions
- arXiv