BUTLER Agent 

Yuchen, the watch wasn't moved to the drawer, it was just picked up (in AI2THOR, held objects hover in front of the camera).

@@comm1234567890 Yeah, I got it. Well, the text says "take watch 1 from drawer 1". Shouldn't it be "from table x"? Thanks!

Oh, I guess I understand. The agent does want to "take watch 1 from drawer 1," but the controller doesn't care "from where". It only uses the mask to complete the action. The agent is just lucky because there is indeed a watch near drawer 1.