mlsanityData to 14 Oct 2026

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

arXiv 2606.20812 ↗First posted 18 Jun 2026cs.LG cs.AIcited by 0

On the map: Diffusion models → Medical imaging and segmentation

Abstract (arXiv)

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 (≈ 307h), 1-2 orders of magnitude (≈ 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 kappa = -0.096). https://jd730.github.io/projects/BFM2

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Medical imaging and segmentation, one of 14 directions in Diffusion models. Builds on 39 of its 66 references that are in the map. First posted on arXiv 18 Jun 2026.

Builds on

39 of 66 references.

Closest ancestors · 2

Same direction, newest first

  1. 2023Open-Vocabulary Panoptic Segmentation with Text-to-Image Diffusion Models8 Mar
  2. 2021Label-Efficient Semantic Segmentation with Diffusion Models6 Dec

Other references · 37

  1. 2017Attention Is All You Needcited by 60,361
  2. 2020Denoising Diffusion Probabilistic Modelscited by 21,076
  3. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,713
  4. 2020A Simple Framework for Contrastive Learning of Visual Representationscited by 12,049
  5. 2023Visual Instruction Tuningcited by 7,869
  6. 2021Masked Autoencoders Are Scalable Vision Learnerscited by 7,403
  7. 2022Scalable Diffusion Models with Transformerscited by 6,054
  8. 2022Flow Matching for Generative Modelingcited by 5,702
  9. 2021Improved Denoising Diffusion Probabilistic Modelscited by 3,761
  10. 2022Elucidating the Design Space of Diffusion-Based Generative Modelscited by 3,045
  11. 2021Efficiently Modeling Long Sequences with Structured State Spacescited by 2,465
  12. 2022Are Transformers Effective for Time Series Forecasting?cited by 1,424
25 more
  1. 2020DiffWave: A Versatile Diffusion Model for Audio Synthesiscited by 1,354
  2. 2017A Closer Look at Spatiotemporal Convolutions for Action Recognitioncited by 1,143
  3. 2023Improving and generalizing flow-based generative models with minibatch optimal transportcited by 879
  4. 2024Chronos: Learning the Language of Time Seriescited by 759
  5. 2024SiT: Exploring Flow and Diffusion-based Generative Models with Scalable Interpolant Transformerscited by 737
  6. 2023Time-LLM: Time Series Forecasting by Reprogramming Large Language Modelscited by 700
  7. 2023A decoder-only foundation model for time-series forecastingcited by 692
  8. 2016EEGNet: A Compact Convolutional Network for EEG-based Brain-Computer Interfacescited by 628
  9. 2024Unified Training of Universal Time Series Forecasting Transformerscited by 614
  10. 2021CSDI: Conditional Score-based Diffusion Models for Probabilistic Time Series Imputationcited by 582
  11. 2023Emergent Correspondence from Image Diffusioncited by 446
  12. 2018Efficient Attention: Attention with Linear Complexitiescited by 368
  13. 2024Large Brain Model for Learning Generic Representations with Tremendous EEG Data in BCIcited by 327
  14. 2021BENDR: using transformers and a contrastive self-supervised learning task to learn from massive amounts of EEG datacited by 199
  15. 2024CBraMod: A Criss-Cross Brain Foundation Model for EEG Decodingcited by 199
  16. 2023SimMTM: A Simple Pre-Training Framework for Masked Time-Series Modelingcited by 115
  17. 2021Transformer-based Spatial-Temporal Feature Learning for EEG Decodingcited by 97
  18. 2023Predict, Refine, Synthesize: Self-Guiding Diffusion Models for Probabilistic Time Series Forecastingcited by 88
  19. 2025REVE: A Foundation Model for EEG -- Adapting to Any Setup with Large-Scale Pretraining on 25,000 Subjectscited by 80
  20. 2022Transformer Convolutional Neural Networks for Automated Artifact Detection in Scalp EEGcited by 55
  21. 2025CSBrain: A Cross-scale Spatiotemporal Brain Foundation Model for EEG Decodingcited by 53
  22. 2024Flow Matching with Gaussian Process Priors for Probabilistic Time Series Forecastingcited by 43
  23. 2020Inter-database validation of a deep learning approach for automatic sleep scoringcited by 38
  24. 2021Self-supervised EEG Representation Learning for Automatic Sleep Stagingcited by 32
  25. 2025CodeBrain: Bridging Decoupled Tokenizer and Multi-Scale Architecture for EEG Foundation Modelcited by 28

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Sensor Geometry as a Flow-Matching Prior for Multi-Channel Brain Signals
  2. 2026TRACE: Temporal Routing with Autoregressive Cross-channel Experts for EEG Representation Learning
  3. 2026MontageFlow: Physically Valid High Fidelity EEG Generation
  4. 2026GRFBrain: Graph-Structured Rectified Flows for EEG Dynamic Modeling
  5. 2025Uni-NTFM: A Unified Foundation Model for EEG Signal Representation Learning
  6. 2026STEAM: A Spatio-TEmporal Alignment Mixture-of-Experts Model with Hierarchical Pre-training for EEG Decoding
  7. 2026NeuroSonic: Conditional Flow Matching for EEG-to-Speech Reconstruction
  8. 2026SpecMoE: Spectral Mixture-of-Experts Foundation Model for Cross-Species EEG Decoding