0 diffusion language-model backbone.
Multiple visual codes let one diffusion model understand and generate images
Intern Lumina U2 represents each visual position with complementary discrete codes, allowing text, images, video and 3D content to share one diffusion backbone and training objective.
Chinese Tech
Qian Yu · Zongyang Qiu · Juncheng Yan · Hongliang Li · Qi Qin · Keqi Wang · +37 more
Shanghai AI Laboratory · Shanghai Jiao Tong University · Huawei Technologies · Dalian University of Technology · Tsinghua University
Research Digest··2 min read
Yu and colleagues built a unified multimodal model that performs visual understanding, image generation and editing using a common discrete representation.
Why this paper
From Shanghai AI Laboratory and 4 others
In one line
A diffusion language model using multi-codebook discrete visual tokens unifies image, video, and 3D understanding with generation and editing in one backbone.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors (3 noted)
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§