作者位次 #2/5ACL 2026Poster· Natural Language Processing
PAR : Training-Free Positional Perturbation and Attention Recycling for Faithful OCR
Yao Yao · Manwen Liao · Weitian Zhang · Zuchao Li · Hai Zhao
摘要
In high-precision scenarios, vision language models suffer from Linguistic Priors Hallucination. When processing familiar text, models tend to over-rely on internal parametric knowledge, effectively “reciting” the content rather than “reading” the image. In this paper, we first systematically inv...