A perspective on behavior as embodied, situated, and hierarchical, and on how physics engines, computer vision, and multimodal language models can support richer quantitative analysis of animal behavior.

An open-source framework for scalable motion imitation learning with physiologically realistic, muscle-actuated humanoids — two validated musculoskeletal embodiments (126- and 416-muscle) plus GPU-parallel training of generalist motor policies.

Our work aims to bridge the gap between images of handwriting and digital ink with a Vision Language Model (PaLI). To our knowledge, this is the first work that effectively does so with arbitrary photos with diverse visual characteristics and backgrounds. Furthermore, it generalizes beyond its training domain and can work on simple sketches. Human evaluation reveals that 87% of the samples produced by our model on the challenging HierText dataset are considered valid tracings of the input image, and 67% look like pen trajectories traced by a human.