african-american-english

This paper audits and mitigates dialect bias in large language models, showing they systematically prefer Standard American English over African American English. The authors introduce activation steering, a training-free method that reduces bias significantly while preserving fluency, and release the largest real-AAE parallel corpus to date.

This paper uses layer-wise probing to investigate how wav2vec 2.0 and Whisper encode consonant cluster reduction in African American English, finding that both models distinguish reduced and canonical forms and preserve cues to underlying stops.