2 papers
cs.CL2026
Guilt by Association: Semantic Leakage Drives Associative Gender Bias
Hazel Chen, Peter West, Hila Gonen
Semantic leakage occurs when a concept in the prompt steers model generation about an entity, even though the concept is irrelevant to it (e.g., a person who likes yellow is said t…
cs.CL2026
Is Convergence Inevitable? Tracing Output Homogeneity Back to Base Models
Alexandrine Fortier, Hazel Chen, Peter West
The lack of diversity in LM content is widely attributed to the alignment process, but how and where exactly in the pipeline this collapse begins is unknown. We argue that output h…