作者位次 #6/6ACL 2026Poster· Natural Language Processing
Selective Test-Time Debiasing for CLIP via Reward Gating
Jaeho Han · Jisoo Yang · Hyeondong Woo · Mingyu Jeon · Sunjae Yoon · Junyeong Kim
摘要
Vision language models (VLMs) demonstrate strong zero-shot performance, but often perpetuate social stereotypes in person-centric queries, yielding skewed demographic distributions. Current debiasing methods apply uniform bias corrections across all input queries regardless of their bias sensitiv...