AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

DC-Gen: Post-Training Diffusion Acceleration with Deeply Compressed Latent Space

Wenkun He, Yuchao Gu, Junyu Chen, Junyi Wu, Wenhang Ge, Dongyun Zou, Yujun Lin, Zhekai Zhang, Haocheng Xi, Muyang Li, Ligeng Zhu, Jincheng YU and 4 more

ECCV 2026 · Image generation › Autoregressive and latent generation · approach: Latent diffusion with learned tokenizers · 6 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
add embedding alignment before fine-tuning, vs direct fine-tuning, when moving to DC-AE-f32 latents (ablation)Wan2.1-T2V-1.3B · video generation training stabilitytraining stability / visual qualityImproved
add embedding alignment vs direct fine-tuning (ablation)FLUX.1-Krea · text-to-image visual qualityimage detailImproved
add patch embedder alignment (ablation)DC-Gen · FLUX.1-Krea / Wan2.1-1.3Bquality and stabilityImproved
add output layer alignment (ablation)DC-Gen · end-to-end fine-tuning convergencesteps to converge: at least double steps without itImproved
use embedding alignment instead of layer-wise LR decay (ablation)FLUX.1-Krea · MJHQ-30K 512FID: 17.83 -> 13.30Improved
use embedding alignment instead of gradual LoRA (ablation)FLUX.1-Krea · MJHQ-30K 512FID: 15.90 -> 13.30Improved
use embedding alignment instead of autoencoder latent space alignment (ablation)FLUX.1-Krea-12B · MJHQ-30K 512gFID: 15.27 -> 13.30Improved
use LoRA instead of full fine-tuning in end-to-end stage (ablation)FLUX.1-Krea · MJHQ-30K 512CLIP Score: 26.85 -> 27.38Improved
use LoRA instead of full fine-tuning in end-to-end stage (ablation)Wan2.1-T2V-1.3B · VBench 480x832Overall: 79.81 -> 83.48Improved
switch to deeply compressed latent (f32) via DC-GenFLUX.1-Krea-12B · MJHQ-30K 1024 GenEvalGenEval and latency: GenEval 0.69 -> 0.71; latency 4.06 -> 0.88 sImproved
switch to deeply compressed latent via DC-GenWan2.1-T2V-14B · VBench 720x1280Overall / latency: 83.73 -> 83.97; 27.52 -> 3.58 minImproved
switch to deeply compressed latent via DC-GenQwen-Image-Edit · GEdit 1K EnglishOverall (O): 7.23 -> 7.20 (tuned vs DC-Gen); latency 57.63 -> 11.96 sNo clear effect

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