Scaling Next-Brain-Token Prediction for MEG
This work addresses the challenge of modeling and generating brain activity data for neuroscience research, representing an incremental advance in applying autoregressive models to MEG data.
The authors tackled the problem of generating long sequences of magnetoencephalography (MEG) data by scaling next-token prediction to handle over 500 hours of MEG data across multiple datasets and scanners, resulting in generations that remained stable over long rollouts and were closer to correct continuations than swapped controls.
We present a large autoregressive model for source-space MEG that scales next-token prediction to long context across datasets and scanners: handling a corpus of over 500 hours and thousands of sessions across the three largest MEG datasets. A modified SEANet-style vector-quantizer reduces multichannel MEG into a flattened token stream on which we train a Qwen2.5-VL backbone from scratch to predict the next brain token and to recursively generate minutes of MEG from up to a minute of context. To evaluate long-horizon generation, we introduce task-matched tests: (i) on-manifold stability via generated-only drift compared to the time-resolved distribution of real sliding windows, and (ii) conditional specificity via correct context versus prompt-swap controls using a neurophysiologically grounded metric set. We train on CamCAN and Omega and run all analyses on held-out MOUS, establishing cross-dataset generalization. Across metrics, generations remain relatively stable over long rollouts and are closer to the correct continuation than swapped controls. Code available at: https://github.com/ricsinaruto/brain-gen.