WorldExam: Benchmarking World Models from Apparent Appearance to Inherent Reactivity
2026-08-03 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors created WorldExam, a new test for checking how well video generation models can act like believable worlds that react naturally to what happens in them. Unlike older tests focusing mostly on video quality or following direct instructions, WorldExam checks if these models understand how scenes should logically respond even if not told explicitly. They tested 20 popular models and found strengths and weaknesses: some are good at controlling the camera, others manage characters well, and some follow complex language commands better. However, none performed strongly in all areas, showing that good video appearance and following instructions don’t ensure realistic world behavior.
controllable video generationworld modelsreactivitybenchmarkcamera controlaction controllanguage-driven modelsvisual qualityinstruction adherencehierarchical evaluation
Authors
Yuxue Yang, Shuyao Shang, Jiahe Wang, Zitong Zhou, Liang Tan, Junhan Zeng, Ruizhi Li, Junyan Li, Yu Liu, Xiao Yang, Yong Li, Jun Zhu, Hongsheng Li, Tieniu Tan, Lue Fan, Zhaoxiang Zhang
Abstract
Controllable video generation models are increasingly being developed as world models. Accordingly, evaluating them in this role extends beyond the apparent appearance of generated videos to the inherent reactivity of the worlds they depict: the ability to infer from the scene state how the world should react and to generate plausible consequences not explicitly described in the input. Yet existing benchmarks mainly assess visual quality or explicit instruction fulfillment by checking whether requested actions and interaction outcomes are realized, leaving inherent reactivity underexamined. We introduce WorldExam, a hierarchical diagnostic benchmark spanning four levels: Visual Quality, Control Adherence, Spatial Consistency, and World Reactivity. It comprises 1,474 cases across eight dedicated tasks and supports unified evaluation of camera-, action-, and language-driven model paradigms. The World Reactivity level evaluates scene-conditioned reactions and goal-directed behaviors beyond what is explicitly specified in the input. Evaluation of 20 representative models reveals a clear capability split. Camera-driven models excel at camera control, but their interfaces do not support dynamic interaction; action-driven models control subjects more precisely but often leave the world unresponsive; and language-driven models perform better on interaction but follow complex controls less faithfully. No model combines broad task coverage with consistently strong performance, showing that high visual quality and explicit instruction fulfillment do not guarantee inherent reactivity.