Perturbational Validity for Foundation Models of Brain Dynamics: A Controlled Proof-of-Principle Simulation
Abstract
Foundation models for human brain recordings are usually evaluated by signal reconstruction, future-state prediction, and transfer to downstream tasks. However, these benchmarks do not establish whether a transferred model remains valid when the system is actively perturbed. We define perturbational validity as the preservation, after limited system-specific adaptation, of the conditional distribution of future trajectories given the current state and a controlled input, and we evaluate it at three levels: time-series accuracy, dynamical-structure similarity, and responses to perturbations excluded from calibration. We demonstrate the framework in an oracle-drift simulation of stochastic bistable systems. Two otherwise identical multilayer perceptrons were trained on drift evaluations from passive or input-driven trajectories. Next, for each held-out system the shared weights were frozen and only a three-dimensional embedding was adapted, with a correctly specified cubic model fitted from scratch as comparator. With two to five system-specific evaluations, perturbational pretraining yielded lower errors in recovering controlled flow, landscape geometry, finite-run occupancy, response distributions, and dose-transition curves. The advantage was reproduced across five independent runs, persisted under full-network adaptation of the passive model, and was attributable to input excitation rather than transition-state coverage: excitation alone lowered controlled-flow error 1.94-fold relative to coverage alone, in five of five runs. The cubic model became competitive as calibration grew, showing that the benefit is specific to few-shot transfer. This controlled demonstration does not test recovery of dynamics from noisy or partially observed brain recordings. It shows why passive prediction should be complemented by prospective evaluation under controlled inputs.
Related
- DCAF: Unifying Data Validity, Evaluation Validity and Predictive Modeling for Single-cell Perturbation Simulation
- Discovery of Bioresorbable Polymer Suture Coatings for Controlled Tissue Regeneration with Multimodal Foundation Models
- How Much Can AI Virtual Cells Gain from Unified Perturbational Modeling?
- Signature Recontextualization: Mapping perturbational signatures across biological contexts
- Validity of Low-Dimensional Representation of Whole-Body Movements
BibTeX
@misc{alanis2026perturbational,
title = {{Perturbational Validity for Foundation Models of Brain Dynamics: A Controlled Proof-of-Principle Simulation}},
author = {José C. Garcí Alanis and Sarah Alizadeh and Marco Rothermel and Bita Shariatpanahi and Mina Kheirkhah and Stefan G. Hofmann and Tim Hahn and Hamidreza Jamalabadi},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.12710},
eprint = {2609.12710},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.12710},
note = {CC BY-SA 4.0},
}
RIS
TY - UNPB
TI - Perturbational Validity for Foundation Models of Brain Dynamics: A Controlled Proof-of-Principle Simulation
AU - José C. Garcí Alanis
AU - Sarah Alizadeh
AU - Marco Rothermel
AU - Bita Shariatpanahi
AU - Mina Kheirkhah
AU - Stefan G. Hofmann
AU - Tim Hahn
AU - Hamidreza Jamalabadi
PY - 2026
DA - 2026/09/11
PB - arXiv
DO - 10.48550/arxiv.2609.12710
UR - https://arxiv.org/abs/2609.12710
AB - Foundation models for human brain recordings are usually evaluated by signal reconstruction, future-state prediction, and transfer to downstream tasks. However, these benchmarks do not establish whether a transferred model remains valid when the system is actively perturbed. We define perturbational validity as the preservation, after limited system-specific adaptation, of the conditional distribution of future trajectories given the current state and a controlled input, and we evaluate it at three levels: time-series accuracy, dynamical-structure similarity, and responses to perturbations excluded from calibration. We demonstrate the framework in an oracle-drift simulation of stochastic bistable systems. Two otherwise identical multilayer perceptrons were trained on drift evaluations from passive or input-driven trajectories. Next, for each held-out system the shared weights were frozen and only a three-dimensional embedding was adapted, with a correctly specified cubic model fitted from scratch as comparator. With two to five system-specific evaluations, perturbational pretraining yielded lower errors in recovering controlled flow, landscape geometry, finite-run occupancy, response distributions, and dose-transition curves. The advantage was reproduced across five independent runs, persisted under full-network adaptation of the passive model, and was attributable to input excitation rather than transition-state coverage: excitation alone lowered controlled-flow error 1.94-fold relative to coverage alone, in five of five runs. The cubic model became competitive as calibration grew, showing that the benefit is specific to few-shot transfer. This controlled demonstration does not test recovery of dynamics from noisy or partially observed brain recordings. It shows why passive prediction should be complemented by prospective evaluation under controlled inputs.
ER -