NewtPhys: Do Foundation Models Understand Newtonian Physics?
For researchers evaluating physics understanding in vision-language and vision foundation models, this dataset provides a more realistic benchmark than existing synthetic ones, revealing current models' shortcomings.
The paper introduces NewtPhys, a 4D physically annotated dataset from real-world scenes with dense annotations including 3D forces and per-pixel quantities. Evaluation of 56 VLMs and 10 VFMs reveals limitations in low-level physics reasoning.
Previous work has evaluated physics reasoning in foundation models using synthetic or semi-synthetic scenes and visual question-answering tasks. However, these benchmarks emphasize high-level events and lack the visual fidelity required to assess true low-level Newtonian understanding. We introduce NewtPhys, a 4D physically annotated dataset built from multiview images of real-world scenes with physics-grounded simulations. The dataset provides dense, fine-grained annotations across timesteps -- including 3D forces and amodal per-pixel quantities covering physics, tracking, semantics and geometry -- bridging the gap between simplistic synthetic setups and realistic visual complexity. Using NewtPhys, we systematically evaluate 56 VLMs, including 54 open-weight models and 2 closed-source frontier models, and 10 VFMs and reveal limitations in low-level physics reasoning. Beyond benchmarking, our dataset enables future research in physics-grounded vision and the development of next-generation physics-aware evaluations. Code and datasets are available at https://astra-vision.github.io/NewtPhys.