RULER-Bench: Probing Rule-based Reasoning Abilities of Next-level Video Generation Models for Vision Foundation Intelligence

Xuming He, Zehao Fan, Hengjia Li, Fan Zhuo, Hankun Xu, Senlin Cheng, Di Weng, Haifeng Liu, Can Ye, Boxi Wu · 2025

Read RULER-Bench: Probing Rule-based Reasoning Abilities of Next-level Video Generation Models for Vision Foundation Intelligence by Xuming He, Zehao Fan, Hengjia Li, Fan Zhuo, Hankun Xu, Senlin Cheng, Di Weng, Haifeng Liu, Can Ye, Boxi Wu (2025) 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-2512-02622.txt.

Description

Xuming He and nine co-authors test whether video generators follow rules, not just look pretty. New video models hold together over time and look impressive, a step toward vision foundation models. Most benchmarks grade beauty, instruction following, and smoothness, which leaves rule based reasoning unchecked.

RULER-Bench probes whether generated videos obey stated physical and logical rules. A model can look right frame by frame while breaking the rule that binds the scene. Testing rules directly helps builders find where video models merely imitate, and where they truly grasp how a scene should unfold. Rules are a stern test: a clip can look smooth and still break the one rule that counts most.

Catalog record

CREATORXuming He, Zehao Fan, Hengjia Li, Fan Zhuo, Hankun Xu, Senlin Cheng, Di Weng, Haifeng Liu, Can Ye, Boxi Wu
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE76 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2512-02622

Browse the full library · About the Tower of Babel