AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

SegQuant: A Semantics-Aware and Generalizable Quantization Framework for Diffusion Models

Jiaji Zhang, Ruichao Sun, Hailiang Zhao, Jiaju Wu, Peng Chen, Hao Li, Yuying Liu, Kingsum Chow, Gang Xiong, Shuiguang Deng

CVPR 2026 · Image generation › Fast diffusion sampling · approach: Training-free diffusion acceleration · 0 citations (OpenAlex)

Read the paper

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 SegLinear (segment-wise quantization from graph structure) (ablation)SmoothQuant on SD3.5 DiT.0.norm1 · SD3.5 single layerFrobenius error: 0.7041 -> 0.5381Improved
add SegLinear (ablation)GPTQ on SD3.5 DiT.11.norm1 context · SD3.5 single layerFrobenius error: 3.0176 -> 1.7637Improved
add SegLinear (ablation)SVDQuant on SD3.5 DiT.0.norm1 · SD3.5 single layerFrobenius error: 0.0861 -> 0.0714Improved
add SegLinear alone (ablation)SD3.5 W8A8 quantization (SmoothQuant + AMax) · MJHQ-30KFID: 23.35 -> 23.36No clear effect
add SegLinear alone (ablation)SD3.5 W8A8 quantization · MJHQ-30KImage Reward: 0.877 -> 0.899Improved
add DualScale alone (ablation)SD3.5 W8A8 quantization · MJHQ-30KFID: 23.35 -> 22.61Improved
combine SegLinear and DualScale (ablation)SD3.5 W8A8 quantization · MJHQ-30KFID / Image Reward: FID 23.35 -> 22.54; IR 0.877 -> 0.952Improved
use SegQuant instead of Smooth+ on FLUX W8A8 intFLUX.1-dev · MJHQ-30KFID: 26.21 -> 22.85Improved
use SegQuant instead of SVDQuant at W4A8SD3.5-DiT · MJHQ-30KFID: 27.95 -> 27.30Improved
use SegQuant instead of Q-Diffusion on SDXL W8A8 fpSDXL-UNet · MJHQ-30KFID: 17.04 -> 17.03No clear effect
add segmentation and dual-scale stepsSD3.5 (RTX 4090) · Inference efficiencyInference time (ms): INT8 278.86 -> SegQuant-8 287.85Hurt

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