The abstract is read at the publisher; the summary is JClub's.
Canonical neural networks implementing active inference adapt to prism-shifted visuomotor environments faster when using pretrained policy matrices, showing rapid error reduction and after-effects after prism removal.
- Why it matters: Animals must recalibrate perception-action coupling when contingencies change suddenly, but the mechanism for flexible generalization to newly encountered environments is unclear.
- What they did: The study builds circuit-level active inference models with multiple policy matrices in parallel and tests an arm-reaching task, comparing pretrained policies for different target positions versus relearning from scratch.
- The result: Pretrained-policy networks adapt substantially faster and reproduce prism adaptation signatures, implying a biologically plausible way to reuse prior sensorimotor knowledge for rapid adjustment.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.