The paper introduces Abra, a family of flow‑matching transformers used to systematically study scaling laws for text‑to‑image diffusion models across three orders of magnitude in compute. It finds that diffusion models scale predictably like language models but need far more data, with compute optimality occurring at roughly 200 image tokens per parameter—ten times the optimal ratio for large language models. The study also shows that diffusion models are robust to overtraining, that more data is preferable to larger models, and that scaling predictability extends to generative quality, optimal CFG settings, representation quality, and training curve shapes.
By Kyle Chickering, Wei-An Lin, Swayam Bhanded, Dan Saunders, Akshat Tripathi, Jiaming Song, Shyamal Buch, Xinchen Yan
arXiv:2610.09450v1 Announce Type: new
Abstract: Pixel-space diffusion models avoid the lossy VAE of latent models, which suggests an advantage on downstream tasks where fine-grained detail matters. W...
By Hanqiu Li Cai (SperidLabs), Chema Garabito (SperidLabs)
Unified image restoration (UIR) aims to recover high-quality (HQ) content from low-quality (LQ) images with different degradations using a single model. Most recent methods adapt large pretrained text-to-image (T2I) latent diffusion models for their strong capacity and generative priors.
arXiv:2605.00809v3 Announce Type: replace
Abstract: In this paper, we present \textbf{Gen}erative \textbf{L}anguage-\textbf{I}mage \textbf{P}re-training (GenLIP), a minimalist generative pretraining...
By Yan Fang, Mengcheng Lan, Zilong Huang, Weixian Lei, Yunqing Zhao, Yujie Zhong, Yingchen Yu, Qi She, Yao Zhao, Yunchao Wei
arXiv:2605. 18324v2 Announce Type: replace-cross Abstract: Representation Autoencoders (RAE) replace traditional VAE with pretrained vision encoders.
By Jaskirat Singh, Boyang Zheng, Zongze Wu, Richard Zhang, Eli Shechtman, Saining Xie
LLaDA-Image is a unified framework that couples a 6B Diffusion Transformer (DiT) trained from scratch with a frozen vision‑language module based on the LLaDA2.0‑Mini diffusion language model. The approach first builds a strong visual generative prior through image‑only pre‑training and mid‑training, then fine‑tunes with a 220M‑sample generation pipeline that includes 98 real images. The resulting model produces highly photorealistic images that accurately follow fine‑grained editing instructions, and a distilled version, LLaDA‑Image‑Turbo, enables fast inference in 2–4 sampling steps. On Qwen‑Image‑Bench, LLaDA‑Image sets new state‑of‑the‑art scores for open‑source models in both English and Chinese tracks, and the authors release weights, code, and detailed recipes to support further research.
By Chuyan Chen, Haoxing Chen, Kun Chen, Zhenglin Cheng, Long Cui, Ruishan Fang, Zhangxuan Gu, Zhicheng Huang, Zhenzhong Lan, Yuanting Lei, Haoquan Li, Jianguo Li, Rongchuan Li, Sidu Li, Tao Lin, Deyuan Liu, Jiacheng Liu, Lin Liu, Yuxuan Lou, Zhisheng Lu, Yuxin Ma, Shuheng Shen, Peng Sun, Chaoyang Wang, Hongjun Wang, Xiaomei Wang, Yongxin Wang, Chengzhang Wu, Hongru Wu, Jun Xie
Diffusion transformer (DiT) research on image generation has converged to a single evaluation setup: class-conditional generation on ImageNet. While methods improve the FID and related metrics, it is increasingly unclear whether they reflect real progress in generative modeling.
Multimodal models often build on architectures designed for generative vision-language modeling, typically combining separately pretrained vision encoders with causal language models. Visual document...
arXiv:2609.05730v1 Announce Type: cross
Abstract: Contrastive Language-Image Pretraining (CLIP) is a building block of many machine learning applications. Scaling laws have guided resource allocation...
By Samir Char, Carles Domingo-Enrich, Randall Balestriero
The paper argues that large-scale text‑to‑image models can be trained effectively on ImageNet, provided the dataset is enriched with carefully crafted text and image augmentations. Using this approach, the authors match the performance of state‑of‑the‑art models such as FLUX, surpassing SD3 on GenEval by +5 points and SDXL on DPGBench by +12, while employing only 1/1000th of the training images and significantly fewer parameters. The method requires just 500 hours of H100 GPU time, making it a more reproducible and accessible alternative to massive web‑scraped datasets.
By L. Degeorge, A. Ghosh, N. Dufour, D. Picard, V. Kalogeiton
arXiv:2609.39222v1 Announce Type: new
Abstract: High-compression tokenizers are essential for scaling latent image generative models. However, aggressive compression creates a fundamental tradeoff be...
By Xu Huang, Ye Huang, Zijun Liao, Yuwei Niu, Xiaojie Li, Menghan Zhou, De Wen Soh, Xiaotong Li, Daquan Zhou
DC-Gen is a post‑training framework that accelerates text‑to‑image diffusion models by using a deeply compressed latent space. It first aligns the base model’s latent representations with a lightweight embedding alignment, then applies minimal LoRA fine‑tuning to preserve generation quality. Experiments on SANA and FLUX.1‑Krea show that DC‑Gen‑FLUX cuts 4K image generation latency by 53× on an NVIDIA H100 and, with NVFP4 SVDQuant, achieves a 138× total speedup on a single NVIDIA 5090 GPU.
By Wenkun He, Yuchao Gu, Junyu Chen, Dongyun Zou, Yujun Lin, Zhekai Zhang, Haocheng Xi, Muyang Li, Ligeng Zhu, Jincheng Yu, Junsong Chen, Enze Xie, Song Han, Han Cai