AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Distillation-free Scaling of Large State-Space Models for Images and Videos

Hamid Suleman, Syed Talal Wasim, Muzammal Naseer, Jüergen Gall

IJCV 2026 · Architectures and efficiency › Attention and transformers · approach: Mamba architectures · 6 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
interleave one Transformer block per seven Mamba blocks (StableMamba) instead of pure Mamba, no distillationVideoMamba-M · ImageNet-1K valTop-1 accuracy (%): 81.4 -> 83.1Improved
replace distilled VideoMamba-B with undistilled StableMamba-BVideoMamba-B · ImageNet-1K valTop-1 accuracy (%): 82.7 -> 83.9Improved
add distillation-initialization to StableMamba-BStableMamba-B · ImageNet-1K valTop-1 accuracy (%): 83.9 -> 84.1Improved
add distillation-initialization to StableMamba-MStableMamba-M · Kinetics-400Top-1 accuracy (%): 82.2 -> 82.5Improved
replace VideoMamba-M (distilled) with StableMamba-M (distilled)VideoMamba-M · Kinetics-400Top-1 accuracy (%): 81.9 -> 82.5Improved
replace VideoMamba-M (distilled) with StableMamba-M (distilled)VideoMamba-M · Something-Something-v2Top-1 accuracy (%): 67.3 -> 68.1Improved
replace VideoMamba-M with StableMamba-MVideoMamba-M · ImageNet-CmCE (%): 51.6 -> 50.5Improved
place the Transformer block in the middle rather than start or end (ablation)StableMamba-T/S · ImageNet-1K valTop-1 accuracy (%)No clear effect
increase Mamba blocks per Transformer block up to 1:7 (ablation)StableMamba · ImageNet-1K valTop-1 accuracy (%): optimal at ratio 1:7Mixed
increase resolution from 224 to 448 (ablation)StableMamba-T · ImageNet-1K valTop-1 accuracy (%): 77.4 -> 79.9Improved
increase frames from 16 to 32 (ablation)StableMamba-T · Kinetics-400Top-1 accuracy (%): 78.6 -> 79.3Improved
use more training data (25% to 100%) (ablation)StableMamba · ImageNet-1K valTop-1 accuracy (%)Improved

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