Let EEG Models Learn EEG
AuthorsYifan Wang, Yijia Ma, Wen Li, Chenyu You
Resources
This paper teaches a transformer to generate EEG as a continuous signal, using flow matching and signal-aware constraints to better preserve the brain-wave patterns that traditional denoising methods often miss.
Key results
JET lowers time-series Fréchet distance versus EEG-GAN on the TUAB benchmark, indicating better spectral-temporal fidelity.
On the TUEV event-recognition benchmark, JET substantially improves generative fidelity over EEG-GAN.
On the TUSZ seizure benchmark, JET achieves a large drop in TS-FID compared with EEG-GAN.
These near-perfect silhouette scores show the generated samples remain well aligned with their conditioning labels across all three benchmarks.
Augmenting the CbraMod classifier with JET-generated data improves held-out accuracy on all three datasets.
The experiments span TUAB, TUEV, and TUSZ from the TUH EEG Corpus, making this a large-scale clinical EEG generation evaluation.
What the paper found
Let EEG Models Learn EEG introduces Just EEG Transformer, or JET, a conditional flow-matching model that generates raw multi-channel EEG as a continuous trajectory rather than through discrete denoising. Built on a Transformer backbone with patch-based tokenization, adaptive layer normalization, and class-balanced sampling, JET learns a vector field that transports Gaussian noise to the EEG distribution while preserving channel identity and long-range temporal dependencies. Its core novelty is a set of structure-preserving constraints: an L1 reconstruction loss motivated by a Laplacian prior, a statistical consistency term matching per-channel mean and standard deviation, and spatiotemporal regularizers using temporal total variation and Pearson correlation. On three TUH EEG Corpus benchmarks—TUAB, TUEV, and TUSZ, totaling more than 10,000 clinical sessions—JET reduces TS-FID by 42% versus EEG-GAN on TUAB (324.18 to 188.27), 47% on TUEV (448.65 to 235.86), and 45% on TUSZ (274.37 to 151.27), while reaching silhouette scores of 0.995, 0.983, and 0.987 and downstream accuracy gains of +2.9%, +3.2%, and +1.7% when augmenting the CbraMod classifier. The paper’s analyses show that JET preserves 1/f spectral scaling, α-band peaks, non-stationary amplitude envelopes, and heavy-tailed amplitude distributions, and its ablations reveal that Gaussian noise initialization is essential: zero-noise training catastrophically degrades TS-FID to over 1,500, confirming that a non-degenerate base distribution is necessary for valid flow-based EEG synthesis.
Original abstract
High-fidelity EEG generation is critical for alleviating data scarcity and addressing privacy constraints in large-scale neural modeling. Despite recent progress, most existing approaches formulate EEG generation via discrete denoising objectives, which inadequately reflect the inherently continuous temporal dynamics and spectral structure of neural activity. As a result, these methods often struggle to preserve long-range temporal dependencies and exhibit mismatches in the spectral and temporal structure of the generated signals. In this work, we argue that effective EEG generation requires models that operate directly on the continuous evolution of neural signals. We introduce Just EEG Transformer (JET), a generative framework based on conditional flow matching that models EEG as raw sequences evolving along continuous trajectories. By learning a smooth vector field that transports noise to the EEG data distribution, JET captures temporal continuity and transient dynamics without relying on discretized denoising schemes or domain-specific representations. To ensure that the learned dynamics remain consistent with key properties of EEG signals, we introduce principled constraints that preserve spectral structure, temporal stationarity, and signal-level statistics. Across three large-scale benchmarks, JET consistently achieves state-of-the-art performance, reducing TS-FID by over 40% compared to strong baselines. Extensive analyses show that JET captures key structural properties of neural dynamics, providing a scalable and principled approach to EEG generation. Project page: https://y-research-sbu.github.io/JET/ .
Read the original paperMore in Generative Models
Browse all 63 papers →RULER: Instance-aware Rubric Rewards for SVG Generation
Hangyu Ran, Yuhao Zheng, Yingying Zhang, Kevin Qinghong Lin, Han Peng
RULER uses instruction-specific visual rubrics as reinforcement-learning rewards to make SVG generation more faithful, stylish, and resistant to reward hacking.
Think Before You Score: Thinking Reward Model for Visual Generation
Xuehai Bai, Zhenchen Tang, Yang Shi, Dianyi Wang, Tengfei Liu, Wanshun Su, Xuanyu Zhu, Ruohui Wang, Haiwen Diao, Haotian Wang, Xiaoling Gu, Yuanxing Zhang
A visual reward model that first decides what matters in each image-generation case, then scores outputs with detailed rubrics to provide better training signals.
WanPE: Towards Cinematic Prompt Enhancement for Modern Text-to-Video Generation
Yubo Zhu, Yawen Shao, Ziyun Dai, Zixun Fang, Kai Zhu, Siyang Sun, Haolan Xue, Chuxin Wang, Tingyu Weng, Jingming Luo, Chen Shi, Lianghua Huang, Yufeng Ai, Yuzheng Wang, Wenyuan Zhang, Yu Shang, Yuxiang Bao, Zoubin Bi, Jie Xiao, Jinbo Xing, Jiaxing Zhao, Chongyang Zhong, Hengjian Chen, Chenwei Xie, Akide Liu, Zhehan Kan, Yu Liu, Wei Zhai, Sheng Zhong, Wei Tong
WanPE turns ordinary text prompts into director-level cinematic plans, substantially improving the quality and consistency of long-form AI-generated videos.