# OOD Detection on Dirty-MNIST vs Fashion MNIST (test)

Source: [OOD Detection on Dirty-MNIST vs Fashion MNIST (test)](<https://www.sota2.com/research/sota/ood-detection-on-dirty-mnist-vs-fashion-mnist-test>)

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

Data updated: 2026-02-26T05:53:14.900Z

Evaluation of Out-of-Distribution (OOD) detection performance using Dirty-MNIST as the in-distribution dataset and Fashion-MNIST as the out-of-distribution dataset, evaluated on the test split.

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 | OOD Detection |
| Dataset (including any stated split) | Dirty-MNIST vs Fashion MNIST (test) |
| Metric | NLL |
| Unit | SCALAR |
| Direction | Lower is better |
| Results loaded | 10 of 10 |
| 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

- [NLL](<https://www.sota2.com/research/sota/ood-detection-on-dirty-mnist-vs-fashion-mnist-test.md?metric=NLL>): Lower is better; unit: SCALAR.

- [Accuracy](<https://www.sota2.com/research/sota/ood-detection-on-dirty-mnist-vs-fashion-mnist-test.md?metric=Accuracy>): Higher is better; unit: PERCENT.

- [ECE](<https://www.sota2.com/research/sota/ood-detection-on-dirty-mnist-vs-fashion-mnist-test.md?metric=ECE>): Lower is better; unit: PERCENT.

- [AUROC (PE)](<https://www.sota2.com/research/sota/ood-detection-on-dirty-mnist-vs-fashion-mnist-test.md?metric=AUROC%20(PE)>): Higher is better; unit: PERCENT.

- [AUROC (MI)](<https://www.sota2.com/research/sota/ood-detection-on-dirty-mnist-vs-fashion-mnist-test.md?metric=AUROC%20(MI)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | NLL | Result date | Report paper |
| --- | --- | --- | --- |
| ENSEMBLE | 0.271 | Not recorded | 68fa75761e209131cea6817b |
| SINGLE | 0.272 | Not recorded | 68fa75761e209131cea6817b |
| ENSEMBLE+OOD HE-CKA | 0.277 | Not recorded | 68fa75761e209131cea6817b |
| DDU | 0.278 | Not recorded | 68fa75761e209131cea6817b |
| HYPERNET | 0.278 | Not recorded | 68fa75761e209131cea6817b |
| SVGD+HE-CKA | 0.298 | Not recorded | 68fa75761e209131cea6817b |
| SVGD+RBF | 0.304 | Not recorded | 68fa75761e209131cea6817b |
| ENSEMBLE+HE-CKA | 0.306 | Not recorded | 68fa75761e209131cea6817b |
| HYPERNET+OOD HE-CKA | 0.325 | Not recorded | 68fa75761e209131cea6817b |
| SVGD+CKApw | 0.377 | Not recorded | 68fa75761e209131cea6817b |

### ENSEMBLE

NLL: 0.271. Result ID: 6993b07e3a6c31484917d792.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75761e209131cea6817b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ede1e209131cea42861.md>)

[Implementation](<https://github.com/ratschlab/repulsive_ensembles>)

Recorded evaluation configuration: {"Backbone":"LeNet","Runs":5}. This record does not assert independent reproduction.

### SINGLE

NLL: 0.272. Result ID: 6993b07e3a6c31484917d791.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75761e209131cea6817b.md>)

Recorded evaluation configuration: {"Backbone":"LeNet","Runs":5}. This record does not assert independent reproduction.

### ENSEMBLE+OOD HE-CKA

NLL: 0.277. Result ID: 6993b07e3a6c31484917d799.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75761e209131cea6817b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75761e209131cea6817b.md>)

[Implementation](<https://github.com/Deep-Machine-Vision/he-cka-ensembles>)

Recorded evaluation configuration: {"Backbone":"LeNet","Runs":5,"OOD Tuning":"OOD Likelihood Minimization"}. This record does not assert independent reproduction.

### DDU

NLL: 0.278. Result ID: 6993b07e3a6c31484917d790.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75761e209131cea6817b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bb91e209131cea2fe02.md>)

Recorded evaluation configuration: {"Backbone":"LeNet","Runs":5,"OOD Metric":"Feature space density (for PE)"}. This record does not assert independent reproduction.

### HYPERNET

NLL: 0.278. Result ID: 6993b07e3a6c31484917d796.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75761e209131cea6817b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa64431e209131cea03950.md>)

Recorded evaluation configuration: {"Backbone":"LeNet","Runs":5}. This record does not assert independent reproduction.

### SVGD+HE-CKA

NLL: 0.298. Result ID: 6993b07e3a6c31484917d795.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75761e209131cea6817b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75761e209131cea6817b.md>)

[Implementation](<https://github.com/Deep-Machine-Vision/he-cka-ensembles>)

Recorded evaluation configuration: {"Backbone":"LeNet","Runs":5,"Kernel":"Cosine similarity feature kernel"}. This record does not assert independent reproduction.

### SVGD+RBF

NLL: 0.304. Result ID: 6993b07e3a6c31484917d793.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75761e209131cea6817b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ede1e209131cea42861.md>)

[Implementation](<https://github.com/ratschlab/repulsive_ensembles>)

Recorded evaluation configuration: {"Backbone":"LeNet","Runs":5,"Kernel":"RBF"}. This record does not assert independent reproduction.

### ENSEMBLE+HE-CKA

NLL: 0.306. Result ID: 6993b07e3a6c31484917d798.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75761e209131cea6817b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75761e209131cea6817b.md>)

[Implementation](<https://github.com/Deep-Machine-Vision/he-cka-ensembles>)

Recorded evaluation configuration: {"Backbone":"LeNet","Runs":5}. This record does not assert independent reproduction.

### HYPERNET+OOD HE-CKA

NLL: 0.325. Result ID: 6993b07e3a6c31484917d797.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75761e209131cea6817b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75761e209131cea6817b.md>)

[Implementation](<https://github.com/Deep-Machine-Vision/he-cka-ensembles>)

Recorded evaluation configuration: {"Backbone":"LeNet","Runs":5,"OOD Tuning":"OOD Likelihood Minimization"}. This record does not assert independent reproduction.

### SVGD+CKApw

NLL: 0.377. Result ID: 6993b07e3a6c31484917d794.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75761e209131cea6817b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa5ee91e209131ce9e6306.md>)

Recorded evaluation configuration: {"Backbone":"LeNet","Runs":5,"Kernel":"Cosine similarity feature kernel"}. This record does not assert independent reproduction.

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