Analysis of HateXplain reveals majority-vote labels mask concentrated annotator disagreement at the hate/offensive boundary, causing models to fail on contested cases with undetected high confidence.
Journal of Artificial Intelligence Research , volume=
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Majority Vote Silences Minority Values: Annotator Disagreement at the Hate/Offensive Boundary in HateXplain
Analysis of HateXplain reveals majority-vote labels mask concentrated annotator disagreement at the hate/offensive boundary, causing models to fail on contested cases with undetected high confidence.