AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

On the Diffusibility of High-Dimensional Latents

Chao Feng, Zhiyang Xu, Bowei Chen, Yuanjun Xiong, Xiyao Wang, Jui-Hsien Wang, Richard Zhang, Zhe Lin, Andrew Owens, Yijun Li

ECCV 2026 · Image generation › Autoregressive and latent generation · approach: High-resolution diffusion synthesis · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
finetune encoder with reconstruction loss (keep v-prediction) (ablation)DINOv2-L RAE · text-to-imageGenEval: 37.63 -> 30.96Hurt
finetune encoder with reconstruction loss (keep v-prediction) (ablation)DINOv2-L RAE · COCO-30kFID: 18.34 -> 29.97Hurt
switch v-prediction to x0-prediction (ablation)FT-DINOv2-L · text-to-imageGenEval: 30.96 -> 39.48Improved
switch v-prediction to x0-prediction (ablation)FT-DINOv2-L · COCO-30kFID: 29.97 -> 16.80Improved
switch v-prediction to x0-prediction (ablation)MAE-B-RAE · text-to-imageGenEval: 36.17 -> 40.89Improved
switch v-prediction to x0-prediction (ablation)MAE-B-RAE · COCO-30kFID: 22.20 -> 17.24Improved
switch v-prediction to x0-prediction (ablation)MAE-B-RAE · DPG-Benchscore: 67.54 -> 73.90Improved
finetune DINOv2-L for reconstruction plus x0-prediction vs frozen DINOv2-L v-predictionDINOv2-L · COCO-30kFID: 18.34 -> 16.80Improved
compress to 32-d latent instead of x0 in 1024-dFT-DINOv2-L · text-to-imageGenEval / FID: GenEval 40.99 vs 39.48; FID 17.64 vs 16.80Mixed
reconstruction finetuning reduces effective dimensionalityDINOv2-L · ImageNet valR90: 672 -> 129Hurt
reconstruction finetuningDINOv2-L · ImageNet-1k linear probeaccuracy (%): 84.5 -> 63.9Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy