# Beyond Accuracy: Robustness, Interpretability and Expressiveness of EEG Foundation Models

arXiv: https://arxiv.org/abs/2605.17562 · HTML: https://mlsanity.com/paper/2605.17562/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Builds on 19 of its 54 references that are in the map.
- First posted on arXiv 17 May 2026.

## Builds on (most cited first, 19 of 19)

- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization](https://mlsanity.com/paper/1610.02391/) (2016)
- [Learning Important Features Through Propagating Activation Differences](https://mlsanity.com/paper/1704.02685/) (2017)
- [SmoothGrad: removing noise by adding noise](https://mlsanity.com/paper/1706.03825/) (2017)
- [Quantifying Attention Flow in Transformers](https://mlsanity.com/paper/2005.00928/) (2020)
- [Towards Faithfully Interpretable NLP Systems: How should we define and evaluate faithfulness?](https://mlsanity.com/paper/2004.03685/) (2020)
- [Large Brain Model for Learning Generic Representations with Tremendous EEG Data in BCI](https://mlsanity.com/paper/2405.18765/) (2024)
- [Generic Attention-model Explainability for Interpreting Bi-Modal and Encoder-Decoder Transformers](https://mlsanity.com/paper/2103.15679/) (2021)
- [CBraMod: A Criss-Cross Brain Foundation Model for EEG Decoding](https://mlsanity.com/paper/2412.07236/) (2024)
- [The Bottom-up Evolution of Representations in the Transformer: A Study with Machine Translation and Language Modeling Objectives](https://mlsanity.com/paper/1909.01380/) (2019)
- [AttnLRP: Attention-Aware Layer-Wise Relevance Propagation for Transformers](https://mlsanity.com/paper/2402.05602/) (2024)
- [What Happens To BERT Embeddings During Fine-tuning?](https://mlsanity.com/paper/2004.14448/) (2020)
- [REVE: A Foundation Model for EEG -- Adapting to Any Setup with Large-Scale Pretraining on 25,000 Subjects](https://mlsanity.com/paper/2510.21585/) (2025)
- [Subject-Aware Contrastive Learning for Biosignals](https://mlsanity.com/paper/2007.04871/) (2020)
- [EEG Foundation Models: A Critical Review of Current Progress and Future Directions](https://mlsanity.com/paper/2507.11783/) (2025)
- [BrainOmni: A Brain Foundation Model for Unified EEG and MEG Signals](https://mlsanity.com/paper/2505.18185/) (2025)
- [Are Large Brainwave Foundation Models Capable Yet? Insights from Fine-tuning](https://mlsanity.com/paper/2507.01196/) (2025)
- [Interpretable and Robust AI in EEG Systems: A Survey](https://mlsanity.com/paper/2304.10755/) (2023)
- [Unmute the Patch Tokens: Rethinking Probing in Multi-Label Audio Classification](https://mlsanity.com/paper/2509.24901/) (2025)

## Built on by (most cited first, 3 of 3)

- [A Negative-Control Protocol for Clinical EEG Foundation-Model Benchmarks: Dataset Identity and External-Cohort Stress Testing](https://mlsanity.com/paper/2607.24519/) (2026)
- [Cross-Cohort Spectral-Temporal Dissociation in Frozen EEG Foundation-Model Representations](https://mlsanity.com/paper/2607.24834/) (2026)
- [Representational and Functional Robustness to Electrode Montages in EEG Foundation Models](https://mlsanity.com/paper/2609.36288/) (2026)

## Abstract (arXiv)

EEG foundation models (EEG-FMs) have been evaluated predominantly on clean, in-distribution accuracy, demonstrating modest gains over supervised baselines and weak frozen representations. This study examines whether these conclusions hold beyond clean accuracy by evaluating six EEG-FMs and a supervised baseline across ten datasets along three layers of analysis: (i) Robustness: we apply test-time perturbations including additive noise, random and region-based channel dropout and region-specific noise injection. Our analyses show that no single model dominates all failure modes. The most noise-robust model is among the most fragile under channel dropout and much of the dropout fragility disappears when channels are removed rather than zero-padded. (ii) Interpretability: using attribution methods in EEG-FMs, we show that models broadly concentrate relevance on task-appropriate brain regions consistent with known neurophysiology. (iii) Expressiveness: we demonstrate that the poor head-only performance previously attributed to low-quality pre-trained representations is largely explained by the pooling strategy and that EEG-FMs possess sufficient representational capacity when their token-level embeddings are preserved. Furthermore, with block-wise probing and attention analysis we show that late blocks are repurposed during fine-tuning, while early blocks already hold task-related information. Our results show that conclusions about EEG-FMs depend on evaluation choices and we recommend that future evaluation of EEG-FMs should report robustness per perturbation type, produce attribution maps and examine multiple pooling strategies.