EEG-Xplain: Decoding Neural Black-Boxes of EEG Foundation Models

• URL: http://arxiv.org/abs/2609.15687v1
• Date: Mon, 14 Sep 2026 14:56:07 GMT
• Title: EEG-Xplain: Decoding Neural Black-Boxes of EEG Foundation Models
• Abstract summary: EEG foundation models such as BIOT, LaBraM, and EEGMamba have achieved remarkable performance in neural signal decoding.<n>We propose a unified attribution framework for interpreting EEG foundation models across heterogeneous architectures.
• Score: 8.430851504111585
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: EEG foundation models such as BIOT, LaBraM, and EEGMamba have achieved remarkable performance in neural signal decoding, but their black-box nature limits clinical trust and neuroscientific validation. We propose a unified attribution framework for interpreting EEG foundation models across heterogeneous architectures. The framework integrates gradient-, perturbation-, and activation-based explanation methods to analyze model behavior in spatial, temporal, and frequency dimensions. Spatially, it identifies critical EEG channels and visualizes their distributions using topographic maps. Temporally, it highlights decision-relevant signal segments through attribution heatmaps. In the frequency domain, it quantifies the contributions of canonical EEG rhythms via spectral perturbation analysis. To assess explanation reliability, we introduce a population-level evaluation combining Area Over the Perturbation Curve (AOPC) and cross-method consistency analysis. The framework further leverages Large Language Models (LLMs) to transform structured attribution outputs into natural-language reports, bridging low-level neural representations and high-level semantic reasoning. Experiments on benchmark datasets, including Mumtaz2016 and TUAB, demonstrate that the generated explanations are consistent with established neurophysiological markers, validating meaningful neural representations while exposing potential dependencies on artifacts and spurious patterns. The proposed framework provides a standardized approach for evaluating the interpretability, reliability, and physiological plausibility of EEG foundation models.

Related papers

• Taming foundation model with invariance-oriented pre-training for broad-spectrum EEG analysis across signal-level, brain-state, and brain-health tasks [46.533103231728724]
  INCEPT is an invariance-oriented EEG foundation model trained on over 11,000 hours of unlabelled clinical EEG.<n>We evaluate INCEPT on a broad-spectrum benchmark of ten datasets spanning three levels of post-acquisition EEG analysis.
  arXiv  Detail & Related papers  (2026-08-25T14:20:02Z)
• Comparing Post-Hoc Explainable AI Methods for Interpreting Black-Box EEG Models in Depression Detection [0.0]
  This study investigates multiple post-hoc explainability methods applied to an InceptionTime architecture trained for EEG-based depression detection.<n>The evaluated methods revealed partially convergent attribution patterns, with recurring emphasis on frontal, temporal, and posterior EEG regions.<n>Overall, the results suggest that different post-hoc explainability approaches capture partially overlapping relevance structures in EEG-based deep learning models for depression detection.
  arXiv  Detail & Related papers  (2026-05-27T18:32:57Z)
• K-Models: a Flexible and Interpretable Method for Ordinal Clustering with Application to Antigen-Antibody Interaction Profiles [65.21921642886367]
  K-Models is a novel framework that integrates ordinal constraints and estimates key underlying elements of the random process generating the observed functional profiles.<n>The proposed method is evaluated through simulations and real-world applications.
  arXiv  Detail & Related papers  (2026-05-14T13:35:44Z)
• A Systematic Survey and Benchmark of Deep Learning for Molecular Property Prediction in the Foundation Model Era [57.5440098776759]
  Molecular property prediction integrates quantum chemistry, cheminformatics, and deep learning to connect molecular structure with biological behavior.<n>This survey traces four complementary paradigms, including Quantum, Descriptor Machine Learning, Geometric Deep Learning, and Foundation Models.<n> Benchmark analyses integrate evidence from both widely used datasets and datasets reflecting industry perspectives.
  arXiv  Detail & Related papers  (2026-04-17T15:16:33Z)
• Stringology-Based Motif Discovery from EEG Signals: an ADHD Case Study [0.0]
  We propose a novel computational framework for analyzing electroencephalography (EEG) time series using methods from stringology.<n>The framework adapts order-preserving matching (OPM) and Cartesian tree matching (CTM) to detect temporal motifs.<n>These findings suggest that ADHD-related EEG alterations involve systematic differences in the structure, stability, and hierarchical organization of recurrent temporal patterns.
  arXiv  Detail & Related papers  (2026-03-03T19:44:55Z)
• Neural Network-based Partial-Linear Single-Index Models for Environmental Mixtures Analysis [10.274230637691717]
  We propose a neural network-based partial-linear single-index (NeuralPLSI) modeling framework.<n>NeuralPLSI bridges semiparametric regression modeling interpretability with the expressive power of deep learning.<n>Our contributions establish NeuralPLSI as a scalable, interpretable, and versatile modeling tool for mixture analysis.
  arXiv  Detail & Related papers  (2025-12-12T14:28:47Z)
• HOLE: Homological Observation of Latent Embeddings for Neural Network Interpretability [3.1865582949300837]
  HOLE (Homological Observation of Latent Embeddings) is a method for analyzing and interpreting deep neural networks through persistent homology.<n>It extracts topological features from neural activations and presents them using a suite of visualization techniques, including Sankey diagrams, heatmaps, dendrograms, and blob graphs.<n>We evaluate HOLE on standard datasets using a range of discriminative models, focusing on representation quality, interpretability across layers, and robustness to input perturbations and model compression.
  arXiv  Detail & Related papers  (2025-12-08T19:20:05Z)
• On the Impact of Graph Neural Networks in Recommender Systems: A Topological Perspective [49.391877616394765]
  In recommender systems, user-item interactions can be modeled as a bipartite graph, where user and item nodes are connected by undirected edges.<n>This graph-based view has motivated the rapid adoption of graph neural networks (GNNs)<n>Despite their empirical success, the reasons why GNNs offer systematic advantages over other approaches remain only partially understood.
  arXiv  Detail & Related papers  (2025-12-08T10:19:43Z)
• WaveNet's Precision in EEG Classification [1.0885910878567457]
  This study introduces a WaveNet-based deep learning model designed to automate the classification of EEG signals into physiological, pathological, artifact, and noise categories.<n>The model was trained, validated, and tested on 209,232 samples with a 70/20/10 percent split.<n>WaveNet's architecture, originally developed for raw audio synthesis, is well suited for EEG data due to its use of dilated causal convolutions and residual connections.
  arXiv  Detail & Related papers  (2025-10-10T09:21:21Z)
• WaveMind: Towards a Conversational EEG Foundation Model Aligned to Textual and Visual Modalities [55.00677513249723]
  EEG signals simultaneously encode both cognitive processes and intrinsic neural states.<n>We map EEG signals and their corresponding modalities into a unified semantic space to achieve generalized interpretation.<n>The resulting model demonstrates robust classification accuracy while supporting flexible, open-ended conversations.
  arXiv  Detail & Related papers  (2025-09-26T06:21:51Z)
• CodeBrain: Towards Decoupled Interpretability and Multi-Scale Architecture for EEG Foundation Model [52.466542039411515]
  EEG foundation models (EFMs) have emerged to address the scalability issues of task-specific models.<n>We present CodeBrain, a two-stage EFM designed to fill this gap.<n>In the first stage, we introduce the TFDual-Tokenizer, which decouples heterogeneous temporal and frequency EEG signals into discrete tokens.<n>In the second stage, we propose the multi-scale EEGSSM architecture, which combines structured global convolution with sliding window attention.
  arXiv  Detail & Related papers  (2025-06-10T17:20:39Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.