When models disagree strongly—measured by high entropy or ensemble variance—it...

https://wiki-global.win/index.php/How_to_Use_Disagreement_to_Catch_Edge_Cases_Before_They_Cause_Harm

When models disagree strongly—measured by high entropy or ensemble variance—it often signals risky or unclear inputs. By flagging the top 1-2% of these uncertain cases, you can route them for expert review or targeted labeling