A Generalist Intracortical Motor Decoder
Abstract
Mapping the relationship between neural activity and motor behavior is a central aim of sensorimotor neuroscience and neurotechnology. While most progress to this end has relied on restricting complexity, the advent of foundation models instead proposes integrating a breadth of data as an alternate avenue for broadly advancing downstream modeling. We quantify this premise for motor decoding from intracortical microelectrode data, pretraining an autoregressive Transformer on 2000 hours of neural population spiking activity paired with diverse motor covariates from over 30 monkeys and humans. The resulting model is broadly useful, benefiting decoding on 8 downstream decoding tasks and generalizing to a variety of neural distribution shifts. However, we also highlight that scaling autoregressive Transformers seems unlikely to resolve limitations stemming from sensor variability and output stereotypy in neural datasets.
Cite
Text
Ye et al. "A Generalist Intracortical Motor Decoder." Advances in Neural Information Processing Systems, 2025.Markdown
[Ye et al. "A Generalist Intracortical Motor Decoder." Advances in Neural Information Processing Systems, 2025.](https://mlanthology.org/neurips/2025/ye2025neurips-generalist/)BibTeX
@inproceedings{ye2025neurips-generalist,
title = {{A Generalist Intracortical Motor Decoder}},
author = {Ye, Joel and Rizzoglio, Fabio and Ma, Xuan and Smoulder, Adam and Mao, Hongwei and Blumenthal, Gary H and Hockeimer, William and Kunigk, Nicolas Guazzelli and Moore, Dalton D. and Marino, Patrick J. and Chowdhury, Raeed H. and Mayo, J. Patrick and Batista, Aaron and Chase, Steven and Boninger, Michael L and Greenspon, Charles M. and Schwartz, Andrew B. and Hatsopoulos, Nicholas G. and Miller, Lee E. and Bouchard, Kristofer and Collinger, Jennifer L and Wehbe, Leila and Gaunt, Robert},
booktitle = {Advances in Neural Information Processing Systems},
year = {2025},
url = {https://mlanthology.org/neurips/2025/ye2025neurips-generalist/}
}