AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Decompose the Distillation: Interpretable Single-Pass Guidance for Diffusion Models

Wuyang Zhang, Xiaolin Xu

NeurIPS 2026 · Image generation › Fast diffusion sampling · approach: Guidance for conditional diffusion · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes an interpretable single-pass guidance method for diffusion models that decomposes distillation of classifier-free guidance, avoiding the 2N forward passes of CFG.

Problem

Classifier-free guidance requires conditional and unconditional evaluations at every sampling step (2N forward passes for N steps); existing single-pass alternatives reduce this cost but the abstract (truncated) suggests they lack interpretability.

Approach

Decomposes the distillation of CFG into interpretable components to obtain guidance in a single forward pass while retaining an adjustable guidance scale; the full mechanism is not stated in the truncated abstract.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Add scale-adaptive modulation to the core student (ablation)CtrlFlow-Core · ImageNet (N=4, s=8)FID-50K: 73.41 -> 63.42Improved
Add time-grid warping to the core student (ablation)CtrlFlow-Core · ImageNet (N=4, s=8)FID-50K: 73.41 -> 59.83Improved
Add both modulation and warping (full model) (ablation)CtrlFlow-Core · ImageNet (N=4, s=8)FID-50K: 73.41 -> 48.95Improved
Use unified composed-output distillation instead of component-wise loss (ablation)CtrlFlow · ImageNet (N=4, s=8)FID-50K: FID degrades by 19.5 when removedImproved
Dual-adapter decomposition alone vs opaque scale-conditioned student (ablation)CtrlFlow-Core vs AGD-MultiScale · ImageNet (N=4, s=8)FID-50K: 52.87 -> 73.41Hurt
Give windowing and dampening to the opaque student as time-varying scale input (ablation)AGD-MultiScale · ImageNet DiT-XL/2FID-50K: 0.7 to 2.3 FID betterImproved
Add alignment gate that needs the v0/g split (ablation)CtrlFlow-Full · ImageNet DiT-XL/2FID-50K: removing it costs 0.6 to 2.4 FIDImproved
Use three-phase curriculum rather than skipping Phase 1 or Phase 2 (ablation)CtrlFlow · ImageNet DiT-XL/2training stability / high-scale FIDImproved
Use rank 64 adapters (ablation)CtrlFlow · ImageNet DiT-XL/2FID / costNo clear effect
Use conditional-field anchor v_cond + (s-1)g (ablation)CtrlFlow-Core · ImageNet DiT-XL/2FID-50K: matches Core; trails Full by 1.1 to 2.6 FIDNo clear effect
Apply windowing and dampening to exact two-pass CFGCFG · ImageNet (N=4, s=8)FID-50K: 53.84 -> 46.02Improved
Distill CFG into single pass with decomposed adapters (vs CFG)CtrlFlow · ImageNet DiT-XL/2, N=25wall-clock latency (ms/img): 843 -> 437Improved

Experimental setup

Baselines: classifier-free guidance

Builds on: classifier-free guidance

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