Shared modeling across stimulation modalities

Determine how to train a shared neural stimulation response model on datasets that mix modalities, particularly when electrical stimulation and optogenetic stimulation produce qualitatively different artifact time courses.

Background

The study evaluates optogenetic stimulation data from two rhesus macaques and two cortical regions. The authors note that extending the approach to electrical stimulation introduces stereotyped stimulation artifacts with temporal profiles that may differ qualitatively from those in optogenetic data. A shared model must therefore accommodate modality-specific response and artifact structure without undermining cross-session generalization.

References

Electrical stimulation, for example, introduces stereotyped artifacts whose temporal shape may itself be well-captured by learned basis functions, but it remains unclear how to train a shared model on datasets that mix modalities with qualitatively different artifact time courses.

— Robust Neural Stimulation Response Modeling Through Meta-Learning and Pretraining  (2608.26649 - Bryan et al., 27 Aug 2026) in Section Discussion, subsection “Limitations,” subsubsection “Dataset scope”

The key open question is whether the shared latent structure of stimulation responses is sufficiently consistent across the substantial biological heterogeneity of such a corpus to support a common prior. We consider this an important open question for the field, and call on experimenters to consider how standardized multi-site stimulation datasets - analogous to FALCON but for stimulation rather than decoding - could be assembled and shared to enable rigorous benchmarking.

— Robust Neural Stimulation Response Modeling Through Meta-Learning and Pretraining  (2608.26649 - Bryan et al., 27 Aug 2026) in Section Discussion, subsection “Towards a foundation model for neural stimulation”