MotionForesight does not take in any language/action input, just the observed video context. The text descriptions go as additional inputs to the baselines where specified.
Ground truthMotionForesightMolmoMotion (lang)MolmoMotion (no lang)Video-gen (lang): generatedVideo-gen (lang): tracksVideo-gen (no lang): generatedVideo-gen (no lang): tracks
taking a mobile phone out of a box
Ground truthMotionForesightMolmoMotion (lang)MolmoMotion (no lang)Video-gen (lang): generatedVideo-gen (lang): tracksVideo-gen (no lang): generatedVideo-gen (no lang): tracks
moving color pencils down
Ground truthMotionForesightMolmoMotion (lang)MolmoMotion (no lang)Video-gen (lang): generatedVideo-gen (lang): tracksVideo-gen (no lang): generatedVideo-gen (no lang): tracks
pretending to throw bulb holder with wire
Ground truthMotionForesightMolmoMotion (lang)MolmoMotion (no lang)Video-gen (lang): generatedVideo-gen (lang): tracksVideo-gen (no lang): generatedVideo-gen (no lang): tracks
opening door
Ground truthMotionForesightMolmoMotion (lang)MolmoMotion (no lang)Video-gen (lang): generatedVideo-gen (lang): tracksVideo-gen (no lang): generatedVideo-gen (no lang): tracks
pushing scissor from right to left
Ground truthMotionForesightMolmoMotion (lang)MolmoMotion (no lang)Video-gen (lang): generatedVideo-gen (lang): tracksVideo-gen (no lang): generatedVideo-gen (no lang): tracks
spinning a disc so it continues spinning
Ground truthMotionForesightMolmoMotion (lang)MolmoMotion (no lang)Video-gen (lang): generatedVideo-gen (lang): tracksVideo-gen (no lang): generatedVideo-gen (no lang): tracks
throwing a pen in the air and letting it fall
Ground truthMotionForesightMolmoMotion (lang)MolmoMotion (no lang)Video-gen (lang): generatedVideo-gen (lang): tracksVideo-gen (no lang): generatedVideo-gen (no lang): tracks
moving mobile phone and a box closer to each other
Ground truthMotionForesightMolmoMotion (lang)MolmoMotion (no lang)Video-gen (lang): generatedVideo-gen (lang): tracksVideo-gen (no lang): generatedVideo-gen (no lang): tracks
pretending to scoop uppuma up with flat ladle with hole
Ground truthMotionForesightMolmoMotion (lang)MolmoMotion (no lang)Video-gen (lang): generatedVideo-gen (lang): tracksVideo-gen (no lang): generatedVideo-gen (no lang): tracks
spinning a bottle that quickly stops spinning
Ground truthMotionForesightMolmoMotion (lang)MolmoMotion (no lang)Video-gen (lang): generatedVideo-gen (lang): tracksVideo-gen (no lang): generatedVideo-gen (no lang): tracks
putting a cup upright on the table
Ground truthMotionForesightMolmoMotion (lang)MolmoMotion (no lang)Video-gen (lang): generatedVideo-gen (lang): tracksVideo-gen (no lang): generatedVideo-gen (no lang): tracks
turning flip flop upside down
Ground truthMotionForesightMolmoMotion (lang)MolmoMotion (no lang)Video-gen (lang): generatedVideo-gen (lang): tracksVideo-gen (no lang): generatedVideo-gen (no lang): tracks
turning eraser upside down
Ground truthMotionForesightMolmoMotion (lang)MolmoMotion (no lang)Video-gen (lang): generatedVideo-gen (lang): tracksVideo-gen (no lang): generatedVideo-gen (no lang): tracks
holding a pencil case
Ground truthMotionForesightMolmoMotion (lang)MolmoMotion (no lang)Video-gen (lang): generatedVideo-gen (lang): tracksVideo-gen (no lang): generatedVideo-gen (no lang): tracks
poking a tire so lightly that it doesn't or almost doesn't move
Ground truthMotionForesightMolmoMotion (lang)MolmoMotion (no lang)Video-gen (lang): generatedVideo-gen (lang): tracksVideo-gen (no lang): generatedVideo-gen (no lang): tracks
pushing quick installation guide book so that it slightly moves
Ground truthMotionForesightMolmoMotion (lang)MolmoMotion (no lang)Video-gen (lang): generatedVideo-gen (lang): tracksVideo-gen (no lang): generatedVideo-gen (no lang): tracks
putting pen onto shoe
Ground truthMotionForesightMolmoMotion (lang)MolmoMotion (no lang)Video-gen (lang): generatedVideo-gen (lang): tracksVideo-gen (no lang): generatedVideo-gen (no lang): tracks
putting trash into trash can
Ground truthMotionForesightMolmoMotion (lang)MolmoMotion (no lang)Video-gen (lang): generatedVideo-gen (lang): tracksVideo-gen (no lang): generatedVideo-gen (no lang): tracks
taking cd from printer
Ground truthMotionForesightMolmoMotion (lang)MolmoMotion (no lang)Video-gen (lang): generatedVideo-gen (lang): tracksVideo-gen (no lang): generatedVideo-gen (no lang): tracks