This paper was posted on arXiv today. It shows surprisingly strong results on ImageNet at 512 resolution (FID 1.41) with one-step generation, while requiring 50% less training-time memory. Do you think this could become the next standard training method for image foundation models? Feel free to leave your comments.
With the rise of vision-language models (VLMs) (such as Qwen-VL and GPT-4.1), new end-to-end OCR models like DeepSeek-OCR have emerged. These models jointly understand visual and textual information, enabling direct interpretation of PDFs without an explicit layout detection step.
However, this paradigm shift raises an important question:
If a VLM can already process both the document images and the query to produce an answer directly, do we still need the intermediate OCR step?