arXiv cs.AI: A latent control model for realistic rodent motion

arXiv cs.AI: A latent control model for realistic rodent motion

The authors present SCAMPER, a system for training an embodied virtual rat to move naturally during both goal-directed behavior and random exploration. They argue that earlier work could train realistic motion through imitation, but did not provide a system that produced natural movement in both settings. SCAMPER combines two priors: a motion prior for control and a behavior prior for exploration. The authors say this combination lets the virtual rat generate realistic exploratory behavior while using task structure effectively. They also report that diverse movements induced by the behavior prior let the system learn from sparse rewards, which they say enables modeling realistic sparse-reward tasks from experimental neuroscience. The proposed system aims to help researchers compare biological and artificial agents by giving them an embodied model that moves more like an animal across different kinds of behavior. The abstract describes these results but does not give the evaluation measures or details needed to assess how realism was judged.

Publication: a72e3104-d914-4706-889e-4e5d5d5ce67a

attribution — pass: The reading consistently attributes the system and its claimed results to the paper's authors ('The authors present', 'They argue', 'The authors say', 'They also report'). It does not present the results as established fact or as Escape Meridian's view. The arXiv cs.AI title prefix matches the publisher.

Here, we introduce Simulated Control via Augmented Motor Priors for Embodied Reinforcement learning (SCAMPER)

summary — pass: The reading leads with the actor and the development: the authors' SCAMPER system for a virtual rat that moves naturally both when pursuing goals and when exploring. It states the gap the authors identify, the two-prior design, the sparse-reward claim and the stated purpose of comparing biological and artificial agents. Technical terms are kept exactly. Term-library tooltips would help general readers with 'prior', 'sparse rewards' and 'embodied model', but the gist is still readable without them. The question is substantive and grounded in the abstract's omission. Omitting the NeurIPS acceptance is not a fidelity problem.

there is no system for embodied animal modeling that produces natural movements both during trained goal-directed behavior, and during exploration driven by random generation

date — pass: originalPublished is 2026-10-19, which matches the v1 submission date. It correctly uses the first submission, not the 3 Nov 2026 revision or the 5 Nov feed date.

Submitted on 19 Oct 2026 (v1), last revised 3 Nov 2026 (this version, v2)

support — pass: Every claim excerpt is an exact substring of the abstract and supports its claim. The stated purpose of helping compare biological and artificial agents is drawn directly from the abstract. The reading does not overstate realism or reliability, and it correctly notes that the abstract gives no evaluation measures. No stakeholder impacts are invented.

Embodied models will enhance our ability to make direct comparisons between biological and artificial agents.

media — pass: The selected media ID is the verified direct link to the arXiv abstract page. It matches the source URL and is the most direct access to the original paper.