B[FM]$^2$: Brain Foundation Model via Flow Matching with SplitUNet

• URL: http://arxiv.org/abs/2606.20812v1
• Date: Thu, 18 Jun 2026 18:01:26 GMT
• Title: B[FM]$^2$: Brain Foundation Model via Flow Matching with SplitUNet
• Abstract summary: Multi-channel EEG signals pose an architectural challenge for flow matching.<n>SplitUNet factorizes each block into separate 1D temporal and 1D electrode convolutions and downsamples only along time.<n>B[FM]$2$ sets a new state of the art on 7 of 9 standard downstream EEG classification tasks.
• Score: 33.19680160939894
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: EEG foundation models can learn generalizable representations from large-scale EEG corpora to enable single-backbone transfer across diverse clinical and brain-computer interface tasks. Existing models typically discretize the continuous multi-channel EEG waveform into patches or codebook tokens and train a transformer with masked self-supervision. Recognizing that this discretization fragments continuous brain rhythms and obscures fine-grained temporal dynamics, we present B[FM]$^2$(Brain Foundation Model via Flow Matching), whose inductive bias aligns with the data by pretraining directly on the raw signal using continuous-time flow matching without patches, tokenization, or masking. However, multi-channel EEG signals pose an architectural challenge for flow matching: time is densely sampled and highly autocorrelated (thousands of timepoints), while the electrode axis is short (tens of channels) at distinct scalp positions. To address this time-electrode asymmetry, we introduce SplitUNet, a velocity network that factorizes each block into separate 1D temporal and 1D electrode convolutions and downsamples only along time, preserving electrode topology throughout the hierarchy. B[FM]$^2$ sets a new state of the art on 7 of 9 standard downstream EEG classification tasks, using a pretraining budget of only 36,895 segments ($\approx$ 307h), 1-2 orders of magnitude ($\approx$ 30x) less than required by existing EEG foundation models. Further, it generates synthetic EEGs that two board-certified neurologists cannot distinguish from brain data (Cohen's $κ=$ -0.096). https://jd730.github.io/projects/BFM2

Related papers

• MUPA$^{2}$E: Multimodal Unified Perception with Asymmetric Attention for Emotion Assessment [4.682152272555128]
  This paper presents MUPAtextsuperscript2E, a unified perception framework that processes facial video and electroencephalography (EEG) through a single shared asymmetric-attention backbone.<n> Facial video is represented through axis-folded frame tokens, while EEG is processed either as a raw multichannel waveform or projected into the spatial domain for multimodal fusion.<n>Using the original recordings, with shorter trials zero-padded to match the longest duration, merged fusion at stride$30$ achieves the highest validation performance and a test accuracy of $70.07%$.
  arXiv  Detail & Related papers  (2026-08-17T01:26:47Z)
• MSBraM: A Multi-scale Self-supervised Brain Foundation Model for Hierarchical EEG Dynamics Learning [16.184230178980705]
  Self-supervised foundation models have recently shown strong potential for electroencephalogram (EEG)-based analysis.<n>Existing approaches struggle to capture the inherently multi-scale temporal structure of EEG signals.<n>We propose MSBraM, a Multi-Scale self-supervised Brain foundation Model designed to learn hierarchical EEG representations.
  arXiv  Detail & Related papers  (2026-07-23T15:03:04Z)
• Dive into Waves: Morlet Spectral Transformer for Cross-Subject Emotion Decoding from EEG [5.886986014593717]
  Cross-subject emotion recognition is a practically important yet challenging problem in brain-computer modeling.<n>Existing approaches rely on large pretrained foundation models, which require massive data yet still struggle with cross-subject variability.<n>We propose the Morlet Spectral Transformer (MST), built around three interfaces and integrated with a EEG-related Transformer backbone.
  arXiv  Detail & Related papers  (2026-05-30T20:38:34Z)
• Torus Graphs for Large Scale Neural Phase Analysis [1.4018975578160686]
  We introduce a score matching procedure that reduces the per-iteration cost to $mathcalO(d2)$, enabling inference on datasets with thousands of variables.<n>This scalable foundation supports analyses of 1,860 frequency-phase features from multi-electrode LFPs and enables two extensions previously inaccessible to TGs or classical circular statistics.
  arXiv  Detail & Related papers  (2026-05-30T02:56:56Z)
• Coupling Models for One-Step Discrete Generation [50.37847093567075]
  Coupling Models is a one-step discrete generative model that learns a direct coupling between discrete sequences and Gaussian latents.<n>It avoids complex continuous flows over the simplex and hand-specified data-to-noise couplings.<n>It reduces LM1B text-generation perplexity by 33% at its lowest-perplexity operating point.
  arXiv  Detail & Related papers  (2026-05-08T03:40:39Z)
• LUNA: Efficient and Topology-Agnostic Foundation Model for EEG Signal Analysis [16.431555478831143]
  We introduce LUNA, a self-supervised foundation model that reconciles disparate electrode geometries.<n>Luna compresses multi-channel EEG into a fixed-size, topology-agnostic latent space.<n>Luna transfers effectively to four downstream tasks: abnormality detection, artifact rejection, slowing classification, and emotion recognition.
  arXiv  Detail & Related papers  (2025-10-25T11:31:27Z)
• 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)
• CEReBrO: Compact Encoder for Representations of Brain Oscillations Using Efficient Alternating Attention [46.47343031985037]
  We introduce a Compact for Representations of Brain Oscillations using alternating attention (CEReBrO)<n>Our tokenization scheme represents EEG signals at a per-channel patch.<n>We propose an alternating attention mechanism that jointly models intra-channel temporal dynamics and inter-channel spatial correlations, achieving 2x speed improvement with 6x less memory required compared to standard self-attention.
  arXiv  Detail & Related papers  (2025-01-18T21:44:38Z)
• Large Brain Model for Learning Generic Representations with Tremendous EEG Data in BCI [6.926908480247951]
  We propose a unified foundation model for EEG called Large Brain Model (LaBraM) LaBraM enables cross-dataset learning by segmenting the EEG signals into EEG channel patches. We then pre-train neural Transformers by predicting the original neural codes for the masked EEG channel patches.
  arXiv  Detail & Related papers  (2024-05-29T05:08:16Z)
• Consistency Trajectory Models: Learning Probability Flow ODE Trajectory of Diffusion [56.38386580040991]
  Consistency Trajectory Model (CTM) is a generalization of Consistency Models (CM) CTM enables the efficient combination of adversarial training and denoising score matching loss to enhance performance. Unlike CM, CTM's access to the score function can streamline the adoption of established controllable/conditional generation methods.
  arXiv  Detail & Related papers  (2023-10-01T05:07:17Z)
• MP-SeizNet: A Multi-Path CNN Bi-LSTM Network for Seizure-Type Classification Using EEG [2.1915057426589746]
  Seizure type identification is essential for the treatment and management of epileptic patients. We present a novel multi-path seizure-type classification deep learning network (MP-SeizNet) MP-SeizNet consists of a convolutional neural network (CNN) and a bidirectional long short-term memory neural network (Bi-LSTM) with an attention mechanism.
  arXiv  Detail & Related papers  (2022-11-09T01:07:20Z)
• Generalizing electrocardiogram delineation: training convolutional neural networks with synthetic data augmentation [63.51064808536065]
  Existing databases for ECG delineation are small, being insufficient in size and in the array of pathological conditions they represent. This article delves has two main contributions. First, a pseudo-synthetic data generation algorithm was developed, based in probabilistically composing ECG traces given "pools" of fundamental segments, as cropped from the original databases, and a set of rules for their arrangement into coherent synthetic traces. Second, two novel segmentation-based loss functions have been developed, which attempt at enforcing the prediction of an exact number of independent structures and at producing closer segmentation boundaries by focusing on a reduced number of samples.
  arXiv  Detail & Related papers  (2021-11-25T10:11:41Z)
• Towards a method to anticipate dark matter signals with deep learning at the LHC [58.720142291102135]
  We study several simplified dark matter (DM) models and their signatures at the LHC using neural networks. We focus on the usual monojet plus missing transverse energy channel, but to train the algorithms we organize the data in 2D histograms instead of event-by-event arrays. This results in a large performance boost to distinguish between standard model (SM) only and SM plus new physics signals.
  arXiv  Detail & Related papers  (2021-05-25T15:38:13Z)

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.