Beyond Static Vision: Scene Dynamic Field Unlocks Intuitive Physics Understanding in Multi-modal Large Language Models

Nanxi Li, Xiang Wang, Yuanjie Chen, Haode Zhang, Hong Li, Yong-Lu Li · 2026

Read Beyond Static Vision: Scene Dynamic Field Unlocks Intuitive Physics Understanding in Multi-modal Large Language Models by Nanxi Li, Xiang Wang, Yuanjie Chen, Haode Zhang, Hong Li, Yong-Lu Li (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-03302.txt.

Description

Li and colleagues probe whether multimodal language models truly grasp intuitive physics, starting with continuum-object dynamics. They introduce two benchmark tasks, next-frame selection and temporal coherence verification, on which even leading models perform poorly. Their Scene Dynamic Field method, which embeds physics simulators in multi-task fine-tuning, gains about 21 percent on fluid tasks and generalizes to unseen domains.

Catalog record

CREATORNanxi Li, Xiang Wang, Yuanjie Chen, Haode Zhang, Hong Li, Yong-Lu Li
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE58992
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.03302

Browse the full library · About the Tower of Babel