arXiv preprint / cs.CV

Iris-3B: Going Beyond the Latent with Pixel-Space Diffusion Training, Conversion and Fine-Tuning

作者：Hanqiu Li Cai (SperidLabs), Chema Garabito (SperidLabs) · 单位：github.com/speridlabs/iris-3b, Chema GarabitoSperidLabs · 会议/期刊：arXiv preprint · 方向：cs.CV · 发布日期：2026-10-08

2026-10-08生成模型
新标签打开 AlphaXiv新标签打开 Code
Iris-3B: Going Beyond the Latent with Pixel-Space Diffusion Training, Conversion and Fine-Tuning figure
AlphaXiv 中文概览（可滚动查看）