An arXiv preprint (2610.27279) describes a latent control model for generating realistic rodent motion. The work is framed within the neuro-AI community's goal of developing embodied models of animal behavior.
The abstract states that embodied models will enhance the ability to make direct comparisons between biological and artificial agents. The proposed model contributes to this effort by focusing on realistic motion as a key component of such comparisons.
Only the abstract was available for this digest, so the article does not cover architecture details or experimental results. No conflicting sources were considered.