| title | Imagined Rollouts are Kinematic, Not Dynamic: A Diagnosis of Long-Horizon World-Model Failure |
|---|---|
| authors | Finn Rasmus Schäfer, Korbinian Moller, Yuan Gao, Christian Oefinger, Sebastian Schmidt, Johannes Betz |
| date | 2026-07-07 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2607.05966 |
| affiliation | TUM - Professorship of Autonomous Vehicle Systems |
| upvotes | 3 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf code |
| primer | World models |
| abstract | Long-horizon failure in world models is conventionally attributed to compounding error, a generic framing that does not distinguish what kind of error compounds. We propose a kinematic-vs-dynamic reframing: world models tend to imagine kinematically rather than dynamically. We operationalize this as the imagined Kinematic-Consistency Error, a per-step diagnostic that measures how far a rollout departs from a closed-form kinematic null, paired with a perturbation protocol that tests whether iKCE responds when physical conditions cross a regime boundary. We instantiate the diagnostic on a released DreamerV3 checkpoint trained on DMC walker-walk, where imagined iKCE runs roughly two orders of magnitude above that of matched real-physics rollouts. Across a friction sweep that crosses the gait-collapse boundary, the model's iKCE stays statistically flat even as the trained policy's reward collapses through the same range, providing the kinematic-not-dynamic signature. The diagnostic distinguishes kinematic from dynamic imagination at horizons longer than the embodiment's gait period. |
cite
@misc{schfer2026imagined,
title={Imagined Rollouts are Kinematic, Not Dynamic: A Diagnosis of Long-Horizon World-Model Failure},
author={Finn Rasmus Schäfer and Korbinian Moller and Yuan Gao and Christian Oefinger and Sebastian Schmidt and Johannes Betz},
year={2026},
eprint={2607.05966},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2607.05966}
}