Metraipolis › archive › dc_460522ca4ef8

RigidBench: Evaluating Rigid-Body Physics in Video Generation Models — arXiv:2608.15555

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.15555
License (as archived)
CC0-1.0
Archived
2026-10-18T19:44:10.802Z
SHA-256 of the archived text
cf9ae7a7536b6a5d37c294b328a40662a29d8b99a47b42f7f9ee7e315adb3798
Stable URL of this page
https://metraipolis.com/documento/dc_460522ca4ef8

How to cite

RigidBench: Evaluating Rigid-Body Physics in Video Generation Models — arXiv:2608.15555 — arXiv (en), https://arxiv.org/abs/2608.15555 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_460522ca4ef8 (Metraipolis archive dc_460522ca4ef8, sha256 cf9ae7a7536b6a5d37c294b328a40662a29d8b99a47b42f7f9ee7e315adb3798)
BibTeX
@misc{dc460522ca4ef8,
  title = {RigidBench: Evaluating Rigid-Body Physics in Video Generation Models — arXiv:2608.15555},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.15555},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_460522ca4ef8; sha256 cf9ae7a7536b6a5d37c294b328a40662a29d8b99a47b42f7f9ee7e315adb3798},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.15555. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Video models are increasingly used to predict what happens next in a scene, yet the metrics commonly used to compare their outputs say little about whether the predicted objects move correctly. Motion, geometry, identity, background stability, and visual similarity can fail independently, but whole-frame scores often mix these errors together. We introduce RigidBench, a simulator-grounded benchmark that compares a generated continuation with a reference rollout from the same initial frame and motion description. Its five rigid-body tasks vary objects, materials, viewpoints, and indoor and outdoor scenes, with per-frame masks, depth, 6-DoF trajectories, and contacts available for scoring. We evaluate eight models on the same 100 examples with ten measurements that keep these aspects separate. The resulting rankings depend strongly on what is measured: no model leads on all ten, and across model means, higher SSIM accompanies larger 3D trajectory error (r = 0.89). RigidBench also includes 5,000 training videos with exact simulator state, which we use to fine-tune and analyze Wan 2.2 TI2V-5B. Full fine-tuning reduces 3D trajectory error by about 20% with almost no change in SSIM, while teacher-forced probes and targeted interventions show that object position is represented throughout Wan's diffusion transformer and used by its denoising computation.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-18 in cs.LG).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2608.15555 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-18T19:02:44.270Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-08-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).