Watch the complete video. Judge the original annotations before suggesting repairs. Review every step and then the four whole-video checks.
1 · Correct
The evidence supports the annotation.
0 · Incorrect
The evidence contradicts it. Explain why.
U · Cannot determine
The view or evidence is insufficient. Do not guess.
For each step
Caption correctness: Is the action actually shown, with the correct object and effect?
Atomicity: Does the caption describe one coherent action? “Fold the paper in half” is one action. “Add rice, stir, and serve” contains several. Small incidental hand movements do not need separate steps.
For the whole video
Check whether the task is correctly described, all essential actions are covered, steps follow the demonstrated order, and the final visible state achieves the task goal. Judge completion from the video itself.
Record the evidence
For an incorrect step, add a short explanation; its step number is recorded automatically. For an incorrect whole-video check, also attach a relevant step number or video time. A truly repeated action is not a duplicated annotation.
Suggested corrections are optional and stay separate from the original text. “Cannot determine” is kept separate from incorrect answers.
Very short actions
Use 0.25× or 0.5× speed and the one-second controls. “Repeat 3 s” loops the preceding three seconds (or the first three seconds at the start). Tap it again to stop repeating. There are no supplied step timestamps, so selecting a caption does not seek the video. Attached times are evidence, not reference timing labels.
Saving your progress
Answers save anonymously as drafts in the shared database. Submit after every step and video check is complete. Return in the same browser to continue your reviews. Clearing browser data or using another device starts a new reviewer session. Review independently; your answers are not shown to other visitors.