Abstract
<title>Abstract</title> <p>Whole-organism simulators turn connectomes and biophysics into testable models of behaviour, but each is governed by thousands of internal parameters normally fixed by invasive physiology— beyond most laboratories and impossible in many species. Here we present PGOB (Parameter Generation by Observed-Behaviour), which obtains these parameters non-invasively, from tracking video, by casting calibration as a behaviour-only inverse problem solved by an interchangeable inference engine. One unchanged formulation generates behaviour-equivalent parameters across four published simulators and two species, rebuilding an entire connectome from a randomised start to its deposited band. Optimised against real video alone, the generated parameters reach the same distance-to-real as the simulator’s invasively-calibrated default—physiological calibration without the physiology. Because a behaviour maps to a set of equivalent parameters, PGOB samples a within-species cohort rather than a single fit, and generation succeeds only within a species, indicating biological constraint—a scalable, non-invasive route reaching species invasive measurement excludes.</p>