LoMoGan

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.

Yeah good point!

Yeah lolll, this is the memory you need

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?

Sounds interesting, will try it out.

Interesting, is this based on an external Vector DB to store and process the PDF?