Despite rapid progress in visual understanding and document transcription, modern vision-language models (VLMs) struggle with a fundamental accessibility tas…
机构:AI4Bharat
来源:arXiv 2610.36195 | AI4Papers 论文推荐平台