From the source
Ai2 released MolmoMotion, a motion forecasting model that predicts future 3D point trajectories from a video frame, query points, and a language instruction.
The model uses Molmo 2 as its backbone and comes in two variants: an autoregressive version (MolmoMotion-AR) and a flow-matching version (MolmoMotion-FM).
Alongside the model, Ai2 published the MolmoMotion-1M dataset of 1.16M videos with 3D point trajectories and action descriptions, and PointMotionBench, a human-validated benchmark of 2.7K video clips for evaluating object-centric 3D motion forecasting.
