AI RESEARCH
CropVLM: Learning to Zoom for Fine-Grained Vision-Language Perception
arXiv CS.AI
•
ArXi:2511.19820v2 Announce Type: replace-cross Vision-Language Models (VLMs) often struggle with tasks that require fine-grained image understanding, such as scene-text recognition or document analysis, due to perception limitations and visual fragmentation. To address these challenges, we