AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

Unveiling the Secret of AdaLN-Zero in Diffusion Transformer

Jie Zhu, Mingyu Ding, Boqiang Duan, Leye Wang, Jingdong Wang

TPAMI 2026 · Image generation › Text-to-image diffusion · approach: Text-driven synthesis and semantic guidance · 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
initialize W_alpha, W_gamma, W_beta with Gaussian std 1e-3 instead of zero (ablation)DiT-XL/2 · ImageNet 256 (50K steps)FID: 78.99 -> 76.21Improved
use too large Gaussian std 5e-2 (ablation)adaLN-Gaussian · ImageNet 256 (50K steps)FID: 78.99 -> 84.37Hurt
use too small Gaussian std 5e-4 (ablation)adaLN-Gaussian · ImageNet 256 (50K steps)FID: 78.99 -> 80.68Hurt
Gaussian init (1e-3) only for W_alpha, W_gamma and W_beta left zero (ablation)adaLN-Zero · ImageNet 256 (50K steps)FID: 78.99 -> 78.62No clear effect
also Gaussian-initialize W_gamma and W_beta (ablation)adaLN-Gaussian W_alpha only · ImageNet 256 (50K steps)FID: 78.62 -> 76.21Improved
replace zero init with adaLN-Gaussian at 400K stepsDiT-XL/2 · ImageNet 256, CFG 1FID: 20.02 -> 17.86Improved
replace zero init with adaLN-Gaussian at 400K steps with CFG 1.5DiT-XL/2 · ImageNet 256, CFG 1.5FID: 6.15 -> 5.28Improved
adaLN-Gaussian at very long training 7000K stepsDiT-XL/2 · ImageNet 256, CFG 1.5FID: 2.27 -> 2.21Improved
adaLN-Gaussian on smaller DiT-B/2DiT-B/2 · ImageNet 256, 400KFID: 42.72 -> 42.55No clear effect
adaLN-Gaussian on DiT-L/2DiT-L/2 · ImageNet 256, 400KFID: 24.40 -> 23.05Improved
adaLN-Gaussian on LlamaVision-XL/2LlamaVision-XL/2 · ImageNet 256, 400KFID: 21.66 -> 20.26Improved
adaLN-Gaussian on text-to-imageDiT-XL/2 · COCO FID-30KFID: 71.41 -> 65.51Improved

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