How self-supervised vision models carve the same image.
Score genomic variants using the GenoLeWM model
Plan robot actions from start and goal images