Synthetic CAD construction videos paired with per-frame atomic action ops, for training video→action (inverse-dynamics) and image→video world models.
v3 extends the v2 dataset (batches b01–b06) with b07–b14. Identical format and contract.
Raw float-millimeter op parameters (no quantization); each clip is translation-canonicalized so the final-mesh bbox is centered at the world origin. The full schema —… See the full description on the dataset page:
https://huggingface.co/datasets/hz6666/3d-wm-atomic-v3.
