ガイド連動クイズ · 中 レベル

Label Bias and Annotator Disagreement Quiz

Examine subjective labels, annotator identity and instructions, proxy targets, disagreement, and safer ways to preserve uncertainty in training data.

関連するガイドパスLabel Bias And Annotator Disagreement
質問 1 の 8

A model learns “toxic” labels that reflect annotators’ assumptions about dialect. What risk is most direct?