2 papers
cs.CL2026
Beyond Black-Box Labels: Interpretable Criteria for Diagnosing Subjective NLP Tasks
Nisrine Rair, Alban Goupil, Valeriu Vrabie +1
Subjective NLP datasets typically aggregate annotator judgments into a single gold label, making it difficult to diagnose whether disagreement reflects unclear criteria, collapsed…
cs.CL2026
When Annotators Disagree, Topology Explains: Mapper, a Topological Tool for Exploring Text Embedding Geometry and Ambiguity
Nisrine Rair, Alban Goupil, Valeriu Vrabie +1
Language models are often evaluated with scalar metrics like accuracy, but such measures fail to capture how models internally represent ambiguity, especially when human annotators…