Distilling Neural Population Geometry into Transformers and GRUs for Motor Control
Abstract
We introduce neuro-distillation (Neuro-Dist), a technique designed to directly transfer computationally useful neural population structure from biological brains into artificial controllers. Using three macaque motor-cortical datasets as neuro-distillation teachers, we align controller internal states with cortical neural population geometry during training; controller deployment requires no neural data. We evaluate gated recurrent unit (GRU) networks and Transformers controlling a simulated arm. Neuro-Dist achieves the lowest median error among five methods, including distillation from movement kinematics, in 10 of 18 comparisons across trained, student-held-out, and new target directions. Seed-split aggregate statistical comparisons significantly favor Neuro-Dist over all six controls on trained and new targets, and five on student-held-out targets (Holm-corrected familywise p<0.05). We furthermore find that Neuro-Dist can substantially improve controller robustness to sensory feedback noise. Mechanistically, Neuro-Dist induced low tangling of trajectories in internal controller state space---a functionally important characteristic of motor cortex, demonstrating successful transfer of a known motor-cortical population geometry feature. These findings establish neuro-distillation as a complement to behavioral distillation and point to a novel class of neuro-interfaces that tap biological neural population structure as a computational resource for improving the control of goal-directed behavior in robotics and embodied AI.
Related
- Saccadic Neural Command Estimation via Forward Dynamics Control: Influence of Extraocular Muscle Geometry on Physiological Plausibility
- Learning continuous updating and control of neural population activities from sparse time samples
- Geometry and cooling effects on an induction self-heating insert for rapid mold temperature control: Experimental, Multiphysics, and artificial neural network-based analysis
- A coarse-to-fine modulation classification scheme utilizing Transformers and Liquid Neural Networks for Smart Grids
- Vision Transformers versus convolutional neural networks for fine-grained orchid genus identification in a species-rich, data-poor flora: a controlled benchmark on the Orchidaceae of New Guinea
BibTeX
@misc{kapitonova2026distilling,
title = {{Distilling Neural Population Geometry into Transformers and GRUs for Motor Control}},
author = {Maryna Kapitonova and Tonio Ball},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.06931},
eprint = {2611.06931},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.06931},
note = {CC BY-SA 4.0},
}
RIS
TY - UNPB
TI - Distilling Neural Population Geometry into Transformers and GRUs for Motor Control
AU - Maryna Kapitonova
AU - Tonio Ball
PY - 2026
DA - 2026/11/01
PB - arXiv
DO - 10.48550/arxiv.2611.06931
UR - https://arxiv.org/abs/2611.06931
AB - We introduce neuro-distillation (Neuro-Dist), a technique designed to directly transfer computationally useful neural population structure from biological brains into artificial controllers. Using three macaque motor-cortical datasets as neuro-distillation teachers, we align controller internal states with cortical neural population geometry during training; controller deployment requires no neural data. We evaluate gated recurrent unit (GRU) networks and Transformers controlling a simulated arm. Neuro-Dist achieves the lowest median error among five methods, including distillation from movement kinematics, in 10 of 18 comparisons across trained, student-held-out, and new target directions. Seed-split aggregate statistical comparisons significantly favor Neuro-Dist over all six controls on trained and new targets, and five on student-held-out targets (Holm-corrected familywise p<0.05). We furthermore find that Neuro-Dist can substantially improve controller robustness to sensory feedback noise. Mechanistically, Neuro-Dist induced low tangling of trajectories in internal controller state space---a functionally important characteristic of motor cortex, demonstrating successful transfer of a known motor-cortical population geometry feature. These findings establish neuro-distillation as a complement to behavioral distillation and point to a novel class of neuro-interfaces that tap biological neural population structure as a computational resource for improving the control of goal-directed behavior in robotics and embodied AI.
ER -