AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

SAF3R: Dynamic Sparse Attention for Feed-Forward 3D Reconstruction Transformers

Jianing Deng, Yuanzhe LI, Jialu Wang, Song Wang, Tianlong Chen, Huanrui Yang, Jingtong Hu

ECCV 2026 · Architectures and efficiency › Attention and transformers · approach: Efficient and robust vision transformers · 1 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
add offline head profiling with tailored kernels (+Dynamic) over uniform sampling (ablation)SAF3R on VGGT · 7ScenesAUC@3: 22.6 -> 24.2 (sparsity 0.85 -> 0.91)Improved
add static kernel for position heads over uniform sampling (ablation)SAF3R on VGGT · 7ScenesAUC@3: 22.6 -> 23.3 (sparsity 0.85 -> 0.90)Improved
add offline head profiling (+Dynamic) over uniform sampling (ablation)SAF3R on DA3 · HiRoomAUC@3: 72.5 -> 81.8Improved
add offline head profiling (+Dynamic) over uniform sampling (ablation)SAF3R on VGGT · HiRoomAUC@30: 82.9 -> 78.8Hurt
add offline head profiling (+Dynamic) over uniform sampling (ablation)SAF3R on VGGT · five datasets (DTU, 7Scenes, ETH3D, ScanNet++)AUC@3: DTU 71.7 -> 74.7; HiRoom 41.2 -> 39.1Mixed
use DINO Top-K (K=4) instead of uniform stride s=32 on correspondence heads (ablation)VGGT layer 14 · Co3D-v2AUC@3: 43.77 -> 48.05Improved
increase DINO Top-K from K=2 to K=4 (ablation)VGGT layer 14 · 7ScenesAUC@3: 22.92 -> 23.34Improved
skip correspondence heads entirely (ablation)VGGT layer 14 · Co3D-v2AUC@30: 90.34 -> 19.55Hurt
increase uniform stride sampling from 1.6% to 3.1% keys (s=64 to s=32) (ablation)VGGT layer 14 · 7ScenesAUC@3: 19.94 -> 22.64Improved
use DINO Top-K (K=4) instead of uniform stride s=32 (ablation)DA3 layer 7 · Co3D-v2AUC@3: 51.74 -> 52.48Improved
SAF3R instead of ADA3 sparse attentionDA3-Giant · ScanNet-50 dense 300 imagesARE: 12.43 -> 5.19Improved
increase K for DINO Top-K on large outdoor scene with loop closures (ablation)DA3 · Oxford Spirestrajectory accuracyImproved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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