Discovering Mechanistic Models of Neural Activity: System Identification in an in Silico Zebrafish
Jan-Matthis Lueckmann, Viren Jain, Michał Januszewski · 2026
Read Discovering Mechanistic Models of Neural Activity: System Identification in an in Silico Zebrafish by Jan-Matthis Lueckmann, Viren Jain, Michał Januszewski (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-04492.txt.
Description
Jan-Matthis Lueckmann, Viren Jain, and Michał Januszewski use a simulated zebrafish to test how well scientists can discover models of brain activity in a 2026 paper. Real neural recordings lack a known ground truth, so model quality is hard to judge.
Their testbed simulates a larval zebrafish, including its body mechanics, so the true system is fully known. A language-model-guided search found predictive models that beat standard forecasting baselines. Yet sensory input alone was not enough for faithful discovery. Models could exploit statistical shortcuts instead. Structural priors, meaning built-in knowledge about the system, proved essential. With them, the search recovered interpretable models that generalized to new conditions.
Catalog record
| CREATOR | Jan-Matthis Lueckmann, Viren Jain, Michał Januszewski |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 73 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-04492