Research map / Detection and segmentation / CNN features
Vision state space models
Apply Mamba-style selective scan models to efficient visual representation learning.
13 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2024–2026) · an approach in CNN features, keyword cluster “vision · imagenet · transformers”
State History Benchmarks Tried Checklist Preprints Who Export
Where it stands
Leveling off Several papers from 2025–26 still report on 1 tracked benchmark, but the best result has barely moved since 2024.
3 papers in 2023–24 and 10 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 0.9% to 2.9%.
What looks settled
- ImageNet-1K val (top-1 accuracy): best up to 2024 84.8, best in 2025–26 85.2 from 2 papers, less than 1% better.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2024 · Vision Mamba: Efficient Visual Representation Learning with Bidirectional State Space Model (ICML 2024 · 417 citations)
- 2025 · Mamba YOLO: A Simple Baseline for Object Detection with State Space Model (AAAI 2025 · 169 citations)
- 2026 · Deformba: Vision State Space Model with Adaptive State Fusion (ICML 2026 · 0 citations)
Earliest papers here
- Vision Mamba: Efficient Visual Representation Learning with Bidirectional State Space Model (ICML 2024 · 417 citations)
- QuadMamba: Learning Quadtree-based Selective Scan for Visual State Space Model (NeurIPS 2024 · 10 citations)
- MambaTree: Tree Topology is All You Need in State Space Model (NeurIPS 2024 · 7 citations)
Benchmark records
The best number each year that beat every earlier one, among papers of this approach reporting their own method. Numbers are extracted by AI from each paper's text; check the paper before citing.
ImageNet-1K val · top-1 accuracy (higher is better)
4 papers, 2024–2025 · small gains only
| Year | Best so far | Set by |
|---|---|---|
| 2024 | 84.8 | MambaTree: Tree Topology is All You Need in State Space Model (NeurIPS 2024 · 7 citations) |
| 2025 | 85.2 | DAMamba: Vision State Space Model with Dynamic Adaptive Scan (NeurIPS 2025 · 3 citations) |
What has been tried
7 papers here report 87 measured changes (a component added, removed or swapped, compared with a baseline): helped 77%, no clear effect 7%, hurt 14%, mixed 2%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Add a backward SSM and backward Conv1d | ImageNet-1K classification ablation · top-1 accuracy | 73.2 -> 73.9 | Helped | Vision Mamba: Efficient Visual Representation Learning with Bidirectional State Space Model (ICML 2024 · 417 citations) |
| Add ODSSBlock | MSCOCO val · AP | 37. 3 -> 43. 1 | Helped | Mamba YOLO: A Simple Baseline for Object Detection with State Space Model (AAAI 2025 · 169 citations) |
| Add deformable branch to forward and backward branches | ImageNet-1K · Top-1 accuracy | 76.9 -> 78.6 | Helped | DefMamba: Deformable Visual State Space Model (CVPR 2025 · 37 citations) |
| replace 4-D scanning with unidirectional 1-D scanning across N/4 channels in four directions | ImageNet-1K · throughput | 803 -> 1125 | Helped | GroupMamba: Efficient Group-Based Visual State Space Model (CVPR 2025 · 22 citations) |
| Replace raster scanning with tree topology scanning | ImageNet-1K · accuracy | 82.6 -> 83.4 | Helped | MambaTree: Tree Topology is All You Need in State Space Model (NeurIPS 2024 · 7 citations) |
| Use dynamic adaptive scanning (DAS) instead of sweeping scan | ImageNet-1K · Top-1 accuracy (%) | 80.8 -> 81.2; 0.4% | Helped | DAMamba: Vision State Space Model with Dynamic Adaptive Scan (NeurIPS 2025 · 3 citations) |
| Add adaptive rotation encoding (ARE) to the RoMA baseline | AID · OA | 69.59 -> 71.70 (TR=20%); 76.80 -> 78.00 (TR=50%) | Helped | RoMA: Scaling up Mamba-based Foundation Models for Remote Sensing (NeurIPS 2025 · 0 citations) |
What reviewers will expect
Datasets and baselines named in the 6 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- ImageNet-1K · 3 of 6 papers
- ADE20K · 3 of 6 papers
- MSCOCO 2017 · 2 of 6 papers
Baselines
- VMamba · 3 of 6 papers
- LocalVMamba · 2 of 6 papers
- Vim · 2 of 6 papers
- Swin Transformer · 2 of 6 papers
- ConvNeXt · 2 of 6 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 13 papers; from OpenAlex and the papers' affiliations).
- Tsinghua University · 2 papers
Use it
Download BibTeX of 13 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Convolution-transformer backbones (101 papers)
- Plain and high-resolution ViTs (69 papers)
- Efficient network training and design (37 papers)
Elsewhere in the topic
- Context modeling for semantic segmentation (241 papers)
- Region-based and query-based detectors (183 papers)
- Local features and context in CNNs (170 papers)
- Context for hard detection conditions (133 papers)
- Domain-specific segmentation and detection (118 papers)
- Box regression and loss design (94 papers)
- Efficient DETR query design (85 papers)
- Boundary-aware instance segmentation (83 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).