Kuis terkait panduan · Sedang Tingkat

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.

Pertanyaan 1 dari 8

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