LightOnOCR-1B: The Case for End-to-End and Efficient Domain-Specific Vision-Language Models for OCR 3 days ago • 45