AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Scalable Feature Matching via State Space Modeling and Sparse Correlation

Sin Wai Choo, Bo Li

CVPR 2026 · Architectures and efficiency › Attention and transformers · approach: Efficient image transformers

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
filter features by L2 norm for sparse coarse correlation versus no filtering (ablation)SLiM · MegaDepthinference time (ms): 128.7 -> 77.0Improved
L2 norm filtering instead of no filtering (ablation)SLiM · MegaDepthAUC@5: 57.9 -> 57.6No clear effect
filter by L2 norm rather than mean (ablation)SLiM · MegaDepthAUC@5: 55.1 -> 57.6Improved
filter by L2 norm rather than variance (ablation)SLiM · MegaDepthAUC@5: 57.2 -> 57.6Improved
mean-based filtering on ScanNet (ablation)SLiM · ScanNetAUC@5: 17.4 -> 0.6Hurt
L2 norm filtering on ScanNet (ablation)SLiM · ScanNetinference time (ms): 41.1 -> 23.9Improved
plug L2-norm filter into LoFTR without retrainingLoFTR · MegaDepthAUC@5 / time: 52.8 -> 53.2; 315.4 ms -> 268.9 msImproved
plug L2-norm filter into Efficient LoFTR without retrainingEfficient LoFTR · MegaDepthAUC@5 / time: 56.4 -> 56.1; 139.2 ms -> 76.0 msMixed
plug L2-norm filter into MatchFormer without retrainingMatchFormer · MegaDepthAUC@5 / time: 53.3 -> 53.4; 555.2 ms -> 487.2 msImproved
reduce recurrent refinement from 4 iterations to 1 (ablation)SLiM · MegaDepthAUC@5: 57.9 -> 57.6; time 77.0 -> 65.3 msHurt
tuned RANSAC inlier thresholdSLiM · MegaDepthAUC@5: 57.9 -> 60.1Improved
SLiM versus semi-dense methods on ScanNet accuracySLiM · ScanNetAUC@5: 18.0 vs 19.6 (ASpanFormer), 19.2 (Efficient LoFTR)Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 3 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