AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

VSSD: Vision Mamba with Non-Causal State Space Duality

Yuheng Shi, Mingjia Li, Minjing Dong, Chang Xu

ICCV 2025 · Architectures and efficiency › Attention and transformers · approach: Mamba architectures · 61 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
replace vanilla SSD with NC-SSD (ablation)VSSD-Micro · ImageNet-1Ktop-1 accuracy (%): 81.0 -> 81.6Improved
replace Bi-SSD with NC-SSD (ablation)VSSD-Micro · ImageNet-1Ktop-1 accuracy (%): 81.4 -> 81.6Improved
replace Bi-SSD with NC-SSD (ablation)VSSD-Micro · ImageNet-1Ktraining throughput (imgs/sec): 399 -> 606Improved
replace vanilla SSD with NC-SSD (ablation)VSSD-Micro · ImageNet-1Ktraining throughput (imgs/sec): 523 -> 606Improved
use standard attention in the last stage (ablation)VSSD-Micro · ImageNet-1Ktop-1 accuracy (%): 81.6 -> 81.8Improved
use overlapped convolutional downsampler instead of patchified (ablation)VSSD-Micro · ImageNet-1Ktop-1 accuracy (%): 81.8 -> 82.5Improved
add weight vector m to NC-SSD (ablation)VSSD-Tiny · ImageNet-1K (100 epochs)top-1 accuracy (%): 32.6 -> 81.8Improved
add weight vector m to NC-SSD (ablation)VSSD-Small · ImageNet-1K (100 epochs)training stability: training crashed in the very first epoch without mImproved
use VSSD-T instead of Swin-TMask R-CNN 1x · COCObox AP: 42.7 -> 46.9Improved
use VSSD-T instead of VMamba-TUperNet · ADE20K single-scalemIoU: 47.3 -> 47.9Improved
use VSSD-T instead of VMambaV9-TImageNet classifier · ImageNet-1Ktop-1 accuracy (%): 82.5 -> 83.7Improved
train with MESAVSSD-T · ImageNet-1Ktop-1 accuracy (%): 83.7 -> 84.1Improved

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