AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

DeltaQuant: 4-bit Video Diffusion Models with Spatiotemporal Delta Smoothing

Xingyang Li, Samuel Tesfai, Zhekai Zhang, Haocheng Xi, Shuo Yang, Lvmin Zhang, Yufei Sun, Kelly Peng, Maneesh Agrawala, Ion Stoica, Kurt Keutzer, Jun-Yan Zhu and 3 more

CVPR 2026 · Image generation › Fast diffusion sampling · approach: Training-free diffusion acceleration

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
quantize delta tokens to 4 bits with FP8 core tokens (DeltaQuant) instead of SVDQuant static smoothingWan2.2 I2V W4A4 · Wan2.2 I2V-A14BPSNR: 21.9 -> 23.2Improved
replace SVDQuant with DeltaQuantLTX-Video T2V W4A4 · LTX-Video T2V-13BLPIPS: 0.193 -> 0.159Improved
replace SVDQuant with DeltaQuantWan2.2 T2V W4A4 · Wan2.2 T2V-A14BVision Reward: 0.083 -> 0.083No clear effect
replace SVDQuant with DeltaQuantWan2.2 T2V W4A4 · Wan2.2 T2V-A14BPSNR: 17.5 -> 17.7Improved
add DeltaQuant on top of Radial and Sage Attention plus LightX2VWan2.2 I2V-A14B · Wan2.2 I2V-A14BLatency (s): 131.0 -> 43.1Improved
add DeltaQuant on top of Radial and Sage Attention plus LightX2VWan2.2 I2V-A14B · Wan2.2 I2V-A14BVision Reward: 0.145 -> 0.148No clear effect
increase cube size to 64 tokens or a single global cube (ablation)DeltaQuant · Wan2.2 (cube size ablation)PSNRHurt
use DeltaQuant instead of raising the SVDQuant low-rank dimension (ablation)Wan2.2 low-rank branch · Wan2.2 (rank ablation)LPIPS vs latencyImproved
increase SVDQuant rank too far (ablation)SVDQuant · Wan2.2 (rank ablation)accuracyHurt
keep core tokens in FP8 instead of BF16 (ablation)DeltaQuant · Wan2.2 (core precision ablation)PSNRNo clear effect
quantize core tokens to NVFP4 instead of FP8 (ablation)DeltaQuant · Wan2.2 (core precision ablation)PSNR: drop of 1.3Hurt
use timestep-aware cube size (16 for 30% of steps, 64 for rest) instead of uniform 32 (ablation)DeltaQuant · Wan2.2 (cube allocation ablation)PSNR: +0.5Improved

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