VideoPhy-2: A Challenging Action-Centric Physical Commonsense Evaluation in Video Generation
This addresses the need for a rigorous benchmark to assess physical commonsense in video generative models, which is crucial for advancing physically-grounded AI simulations, though it is incremental as it builds on prior evaluation efforts.
The paper tackles the problem of evaluating physical commonsense in video generation by introducing VideoPhy-2, a dataset with 200 diverse actions, and finds that even the best model achieves only 22% joint performance on the hard subset, revealing major shortcomings in adherence to physical rules like conservation laws.
Large-scale video generative models, capable of creating realistic videos of diverse visual concepts, are strong candidates for general-purpose physical world simulators. However, their adherence to physical commonsense across real-world actions remains unclear (e.g., playing tennis, backflip). Existing benchmarks suffer from limitations such as limited size, lack of human evaluation, sim-to-real gaps, and absence of fine-grained physical rule analysis. To address this, we introduce VideoPhy-2, an action-centric dataset for evaluating physical commonsense in generated videos. We curate 200 diverse actions and detailed prompts for video synthesis from modern generative models. We perform human evaluation that assesses semantic adherence, physical commonsense, and grounding of physical rules in the generated videos. Our findings reveal major shortcomings, with even the best model achieving only 22% joint performance (i.e., high semantic and physical commonsense adherence) on the hard subset of VideoPhy-2. We find that the models particularly struggle with conservation laws like mass and momentum. Finally, we also train VideoPhy-AutoEval, an automatic evaluator for fast, reliable assessment on our dataset. Overall, VideoPhy-2 serves as a rigorous benchmark, exposing critical gaps in video generative models and guiding future research in physically-grounded video generation. The data and code is available at https://videophy2.github.io/.