EQA: embodied QA

EQA asks you to watch a robot or person perform a task and answer a question about the video. Prompts may ask for counts, object names, action order, or whether an event occurred.

Method

1. Watch the whole video and note the initial scene and object positions.
2. Recheck relevant moments to count events or confirm order.
3. Use only facts supported by the video. Do not give an exact count when objects are hidden or indistinguishable.
4. Follow the requested format exactly. Do not add explanations when the prompt asks for a number or list only.

Review

Check the whole clip, repeated pick-ups or movements, and every object name, color, and sequence against visible evidence.