Research map / Large language models / Instruction and fine-tuning
EEG-FM-Bench: A Comprehensive Benchmark for the Systematic Evaluation and Diagnostic Analyses of EEG Foundation Models
Wei Xiong, Jiangtong Li, Jie Li, Kun Zhu, Changjun Jiang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| fine-tune multi-task instead of single-task (ablation) | EEG foundation models · EEG-FM-Bench (14 datasets) | balanced accuracy | Mixed |
| freeze the backbone instead of full-parameter fine-tuning (ablation) | EEG foundation models · EEG-FM-Bench (14 datasets) | balanced accuracy | Hurt |
| use LoRA instead of full-parameter fine-tuning (ablation) | CBraMod and REVE vs weaker models · EEG-FM-Bench (14 datasets) | balanced accuracy | Mixed |
| initialize from pre-trained checkpoints instead of scratch (ablation) | EEG foundation models · EEG-FM-Bench (14 datasets) | balanced accuracy | Improved |
| use a large compressing MLP head instead of average pooling (ablation) | classifier head · Motor imagery (PhysioMI, BCIC-2A) | balanced accuracy | Improved |
| use attention pooling or large MLP head instead of average pooling (ablation) | classifier head · non-motor-imagery tasks | balanced accuracy | Hurt |
| use general time-series models (Mantis, Moment) instead of EEG-FMs (ablation) | time-series FMs · EEG-FM-Bench (14 datasets) | balanced accuracy | Mixed |
| scale model size and pre-training data | EEG foundation models · EEG-FM-Bench (14 datasets) | transfer performance | No clear effect |
| pre-train with masked reconstruction objective | EEG foundation models · gradient analysis on downstream classification | gradient alignment | Hurt |
| fine-tune CSBrain from scratch (ablation) | CSBrain · EEG-FM-Bench (14 datasets) | training stability | Hurt |
Closest papers by meaning
- NeuroAtlas: Benchmarking Foundation Models for Clinical EEG and Brain-Computer Interfaces (NeurIPS 2026 · similarity 0.93)
- EEG Benchmarking Needs a Task Specification Layer: NeuroDoc for Rulebook-Guided, Executable Benchmark Construction (NeurIPS 2026 · similarity 0.92)
- What do EEG Foundation Models Capture from Human Brain Signals? (NeurIPS 2026 · similarity 0.91)
- BrainOmni: A Brain Foundation Model for Unified EEG and MEG Signals (NeurIPS 2025 · similarity 0.89)
- CBraMod: A Criss-Cross Brain Foundation Model for EEG Decoding (ICLR 2025 · similarity 0.88)
- Brain-OF: An Omnifunctional Foundation Model for fMRI, EEG and MEG (NeurIPS 2026 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
