arXiv Computer Vision
World Models' Last Exam in Physics
Read the original on arXiv Computer Vision →The Flow has not summarised this story yet — read it at arXiv Computer Vision.
Related stories
arXiv AI
Aug 7
GAUGE: A Measurement-Grounded Benchmark for Physical Fidelity in Simulation Engines and Video World Models
arXiv:2608. 05948v1 Announce Type: new Abstract: Physics engines facilitate large-scale training and evaluation for embodied intelligence, while generative video world models are emerging as implicit simulators of future states and interactions.
Hugging Face Trending Papers
Aug 6
GAUGE: A Measurement-Grounded Benchmark for Physical Fidelity in Simulation Engines and Video World Models
Physics engines facilitate large-scale training and evaluation for embodied intelligence, while generative video world models are emerging as implicit simulators of future states and interactions. However, existing evaluations of physical fidelity are often conducted in isolation and rely heavily on perceptual similarity or human judgments, providing limited insight into which physical principles or parameters are violated.
arXiv Computer Vision
Sep 22
HappyWorld-Bench
arXiv:2609.24308v1 Announce Type: new Abstract: Evaluating world models requires assessing both the quality of the worlds they generate and their consistency and responsiveness under exploration, int...
Hugging Face Trending Papers
Aug 3
PhyCheck: Fine-Grained Evidence-Grounded Dataset for Physical Law Understanding in Video-LLMs
Embodied intelligence and world models require video understanding systems to go beyond recognizing objects and actions and develop an understanding of physical regularities. However, despite their strong performance on general video understanding tasks, current video-language models still struggle to reliably determine whether an observed event conforms to specific physical laws.
arXiv AI
Aug 5
PhyCheck: Fine-Grained Evidence-Grounded Dataset for Physical Law Understanding in Video-LLMs
arXiv:2608. 02150v2 Announce Type: replace-cross Abstract: Embodied intelligence and world models require video understanding systems to go beyond recognizing objects and actions and develop an understanding of physical regularities.
arXiv Computer Vision
6d ago
PhysicsLENS: Diagnosing Physical Property Blindness in Video Generation Models
PhysicsLENS is a new dataset and benchmark designed to evaluate how well video generation models capture physical properties relevant to robotics. It consists of matched scenario pairs that share the same conditioning frame and task but differ in underlying physics, and it covers seven physical domains such as collision, gravity, and friction. The benchmark includes over 400 human-annotated labels from four video generation models, revealing that many plausible-looking videos still ignore the specified physical property.