AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Latent Forcing: Reordering the Diffusion Trajectory for Pixel-Space Image Generation

Alan Baade, Eric Ryan Chan, Kyle Sargent, Changan Chen, Justin Johnson, Ehsan Adeli, Li Fei-Fei

ICML 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
denoise DINOv2 latents earlier than pixels (alpha shift 16 vs 1/64) (ablation)Multi-Schedule LF-DiT, DINOv2-B+Reg · ImageNet FID-10K without guidanceFID: 55.35 -> 18.65Improved
order Data2Vec2 latents before pixels (alpha shift 16 vs 1/64) (ablation)Multi-Schedule LF-DiT, Data2Vec2 · ImageNet FID-10K without guidanceFID: 55.19 -> 23.61Improved
order 64x64 downsampled pixels before full-resolution pixels (alpha 16 vs 1/64) (ablation)Multi-Schedule LF-DiT · ImageNet FID-10K without guidanceFID: 44.51 -> 42.35Improved
replace JiT+REPA distillation with Latent Forcing DINOv2 orderingJiT, conditional ImageNet 80 epochs · ImageNet 256 FID-50K unguidedFID: 18.60 -> 9.76Improved
replace JiT+REPA with Latent Forcing in unconditional generationJiT, unconditional ImageNet 80 epochs · ImageNet 256 FID-50K guidedFID: 24.40 -> 13.36Improved
use cascaded schedule instead of variance shift or linear offset (ablation)Single-Schedule LF-DiT-L · ImageNet FID-10K unguidedFID: 13.48 -> 12.42 (var shift); 20.98 -> 12.42 (linear offset)Improved
add output expert layers (split last 4 layers into latent and pixel experts) (ablation)LF-DiT-L · ImageNet FID-10K unguidedFID: 12.93 -> 12.42Improved
sample 10% of pixel timesteps from U[0,0.5] instead of logit-normal (ablation)LF-DiT-L · ImageNet FID-10K unguidedFID: 12.98 -> 12.42Improved
add up to 25% noise to latents during pixel timesteps in training (ablation)LF-DiT, 200 epochs · ImageNet FID-10K unguidedFID: 16.47 -> 10.93Improved
add up to 50% training noise to latents during pixel timesteps (80 epochs) (ablation)LF-DiT · ImageNet FID-10K unguidedFID: 13.48 -> 14.64Hurt
add noise to latents at inference (ablation)LF-DiT, 200 epochs, trained with 25% noise · ImageNet FID-10KFID: 10.93 -> 12.55Hurt
raise probability of sampling latent timestep p_latent from 0.4 to 0.8 (ablation)LF-DiT cascaded training · ImageNet FID-10K without guidanceFID: 12.42 -> 16.13Hurt

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