NUS's Jeveration turns a frozen vision-language model into a single-pass image generator

NUS-DSA3101 · hf · 2026-11-19

Researchers at the National University of Singapore present Jeveration, showing that "System One models" can generate too: it repurposes a frozen vision-language model to produce coherent images and perform image-to-image and dense prediction tasks in a single forward pass, with fast 512×512 generation.

The work bypasses the multi-step denoising of diffusion models, exploring whether understanding-oriented models can directly take on generation.

Original post →

More from Multimodal

Multimodal channel →