Bias in Annotation Instructions
Mor Geva discusses their research on bias in annotation instructions and how it impacts natural language understanding datasets. They highlight the origin of bias in the data set creation process and the patterns observed in the examples and instructions given to annotators. The conversation explores the implications of this bias and raises questions about the annotators' understanding and the clarity of instructions.In this clip
From this podcast
Data Skeptic
StrategyQA and Big Bench
Related Questions