LightOnOCR-1B: The Case for End-to-End and Efficient Domain-Specific Vision-Language Models for OCR 5 days ago • 48