Metraipolis › archive › dc_adb56fc2ce9c
PAWBench: How Far Are We from Probabilistically Aligned World Modeling? — arXiv:2608.27345
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.27345
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-17T14:30:49.885Z
- SHA-256 of the archived text
c740ade8dad212c88942e3f6680ffc304e582269e30e52cef0a861e050e76bc5- Stable URL of this page
https://metraipolis.com/documento/dc_adb56fc2ce9c
How to cite
PAWBench: How Far Are We from Probabilistically Aligned World Modeling? — arXiv:2608.27345 — arXiv (en), https://arxiv.org/abs/2608.27345 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_adb56fc2ce9c (Metraipolis archive dc_adb56fc2ce9c, sha256 c740ade8dad212c88942e3f6680ffc304e582269e30e52cef0a861e050e76bc5)
BibTeX
@misc{dcadb56fc2ce9c,
title = {PAWBench: How Far Are We from Probabilistically Aligned World Modeling? — arXiv:2608.27345},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.27345},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_adb56fc2ce9c; sha256 c740ade8dad212c88942e3f6680ffc304e582269e30e52cef0a861e050e76bc5},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.27345. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Recent video generation models are increasingly framed as world models. Many physical processes can unfold in more than one valid way. Therefore, a world model should reproduce not only a plausible trajectory, but also the distribution of possible behaviors under the same initial observation and action. We call this distribution-level requirement probabilistic alignment. However, existing evaluations largely assess individual-video plausibility and do not test whether repeated generations recover the correct distribution. This raises a central question: how far are current video generators from probabilistically aligned world modeling? To answer it, we formalize probabilistic alignment as a distributional criterion for world models and introduce PAWBench, a benchmark for evaluating video generators as stochastic samplers of world dynamics. We further introduce PAWEval, an outcome-level protocol that converts repeated video rollouts into empirical distributions over possible physical behaviors. Across 50 scenarios and eleven current systems, no model consistently matches the reference probabilities while recovering the range of valid behaviors. Having established this gap, we test whether language prompts, initial noise sampling, or model training can reshape the model's predictive distribution. We believe our work can serve as a foundation for future efforts to move towards probabilistically aligned world modeling. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-28 in cs.AI). Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2608.27345 — 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 2026-10-17T13:16:53.707Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-28?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).