The label aggregation score is very well designed, and the paper well-written. I do have some serious concerns regarding the involvement of AI votes, especially regarding the possible model collapse, when the AI greatly influences its own training data.
The whole review is appended.

The authors have eloquently addressed the concerns and amended them in their work. This reviewer has full confidence that the authors are aware of the possible issues in application and know ways to track and amend them, should it ever come up.