The Penn Computer Graphics Lab works at the intersection of computer graphics, computer vision, AI, and robotics, with the goal of advancing how we reconstruct and represent the 3D world and generate new virtual worlds. Building on our foundational contributions to neural scene representations and rendering, including efficient sparse representations, neural surface reconstruction, and 3D generative models, we have recently expanded our research to physics-grounded 3D/4D reconstruction and physics-based character animation. By bringing physics into 3D vision and generative models, we aim to create digital humans and environments that are not only visually compelling but also physically plausible and interactive, with applications in robotics. Representative projects include PhysCtrl, PhysHMR, Pixie, ModSkill, and Vid2Sim.