# Sparse Dictionary Learning on ImageNet-1k DINOv2-B CLS embeddings (train)

Source: [Sparse Dictionary Learning on ImageNet-1k DINOv2-B CLS embeddings (train)](<https://www.sota2.com/research/sota/sparse-dictionary-learning-on-imagenet-1k-dinov2-b-cls-embeddings-train>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-04-30T06:52:24.888Z

Evaluation of Sparse Dictionary Learning performance using pre-computed DINOv2-B CLS embeddings extracted from the training split of ImageNet-1k, measured by reconstruction (Recon), sparsity probability (SpProb), and auto-interpolation (AutoInterp) metrics.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Sparse Dictionary Learning |
| Dataset (including any stated split) | ImageNet-1k DINOv2-B CLS embeddings (train) |
| Metric | Reconstruction Score |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 2 of 2 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Reconstruction Score](<https://www.sota2.com/research/sota/sparse-dictionary-learning-on-imagenet-1k-dinov2-b-cls-embeddings-train.md?metric=Reconstruction%20Score>): Higher is better; unit: PERCENT.

- [Sparsity Probability](<https://www.sota2.com/research/sota/sparse-dictionary-learning-on-imagenet-1k-dinov2-b-cls-embeddings-train.md?metric=Sparsity%20Probability>): Higher is better; unit: PERCENT.

- [Auto-Interpolation Score](<https://www.sota2.com/research/sota/sparse-dictionary-learning-on-imagenet-1k-dinov2-b-cls-embeddings-train.md?metric=Auto-Interpolation%20Score>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Reconstruction Score | Result date | Report paper |
| --- | --- | --- | --- |
| DB-KSVD | 81.7 | Not recorded | 68fa78c91e209131cea7a815 |
| SAE | 79.4 | Not recorded | 68fa78c91e209131cea7a815 |

### DB-KSVD

Reconstruction Score: 81.7. Result ID: 69f2fb5b47f34ce924d8ce42.

[Reporting paper](<https://www.sota2.com/research/paper/68fa78c91e209131cea7a815.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78c91e209131cea7a815.md>)

Recorded evaluation configuration: {"Sparsity (k)":"32","Sparse Probing Sparsity (ksp)":"1","Dictionary Size (m)":"4096","Backbone Model":"DINOv2-B"}. This record does not assert independent reproduction.

### SAE

Reconstruction Score: 79.4. Result ID: 69f2fb5b47f34ce924d8ce41.

[Reporting paper](<https://www.sota2.com/research/paper/68fa78c91e209131cea7a815.md>)

Recorded evaluation configuration: {"Sparsity (k)":"32","Sparse Probing Sparsity (ksp)":"1","Dictionary Size (m)":"4096","Backbone Model":"DINOv2-B"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.