# Sequential Distribution Shift Detection on ImageNet-C, ImageNet-R, and Office-Home streams

Source: [Sequential Distribution Shift Detection on ImageNet-C, ImageNet-R, and Office-Home streams](<https://www.sota2.com/research/sota/sequential-distribution-shift-detection-on-imagenet-c-imagenet-r-and-office-home-streams>)

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

Data updated: 2026-06-01T09:15:23.412Z

Evaluation of model performance in detecting sequential distribution shifts across ImageNet-C, ImageNet-R, and Office-Home datasets, measured by metrics such as detection delay and false alarm rate.

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 | Sequential Distribution Shift Detection |
| Dataset (including any stated split) | ImageNet-C, ImageNet-R, and Office-Home sequential streams |
| Metric | Detection Delay (samples) |
| Unit | TIME |
| Direction | Lower is better |
| Results loaded | 6 of 6 |
| 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

- [Detection Delay (samples)](<https://www.sota2.com/research/sota/sequential-distribution-shift-detection-on-imagenet-c-imagenet-r-and-office-home-streams.md?metric=Detection%20Delay%20(samples)>): Lower is better; unit: TIME.

- [FAR (%)](<https://www.sota2.com/research/sota/sequential-distribution-shift-detection-on-imagenet-c-imagenet-r-and-office-home-streams.md?metric=FAR%20(%25)>): Lower is better; unit: PERCENT.

## Recorded results

| Method | Detection Delay (samples) | Result date | Report paper |
| --- | --- | --- | --- |
| E-SHIFT (ours) | 17.2 | Not recorded | 68fa7bde1e209131cea8ae4f |
| E-SHIFT (no bootstrap) | 17.8 | Not recorded | 68fa7bde1e209131cea8ae4f |
| SeqMMD-E | 20.3 | Not recorded | 68fa7bde1e209131cea8ae4f |
| Shiryaev–Roberts | 22.7 | Not recorded | 68fa7bde1e209131cea8ae4f |
| CUSUM | 26.3 | Not recorded | 68fa7bde1e209131cea8ae4f |
| Fixed-MMD | 44.1 | Not recorded | 68fa7bde1e209131cea8ae4f |

### E-SHIFT (ours)

Detection Delay (samples): 17.2. Result ID: 6a1d4d92fbc373592c16c4b9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bde1e209131cea8ae4f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7bde1e209131cea8ae4f.md>)

Recorded evaluation configuration: {"Threshold (τ)":"200","correction":"bootstrap corrected ψ¯(λ)"}. This record does not assert independent reproduction.

### E-SHIFT (no bootstrap)

Detection Delay (samples): 17.8. Result ID: 6a1d4d92fbc373592c16c4b8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bde1e209131cea8ae4f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7bde1e209131cea8ae4f.md>)

Recorded evaluation configuration: {"Threshold (τ)":"200","correction":"plug-in ψˆ(λ)"}. This record does not assert independent reproduction.

### SeqMMD-E

Detection Delay (samples): 20.3. Result ID: 6a1d4d92fbc373592c16c4b7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bde1e209131cea8ae4f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa5e131e209131ce9df11b.md>)

Recorded evaluation configuration: {"Threshold (τ)":"200","Backbone":"CLIP"}. This record does not assert independent reproduction.

### Shiryaev–Roberts

Detection Delay (samples): 22.7. Result ID: 6a1d4d92fbc373592c16c4b6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bde1e209131cea8ae4f.md>)

Recorded evaluation configuration: {"Threshold (τ)":"200"}. This record does not assert independent reproduction.

### CUSUM

Detection Delay (samples): 26.3. Result ID: 6a1d4d92fbc373592c16c4b5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bde1e209131cea8ae4f.md>)

Recorded evaluation configuration: {"Threshold (τ)":"200"}. This record does not assert independent reproduction.

### Fixed-MMD

Detection Delay (samples): 44.1. Result ID: 6a1d4d92fbc373592c16c4b4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bde1e209131cea8ae4f.md>)

Recorded evaluation configuration: {"Threshold (τ)":"200","window size":"100"}. This record does not assert independent reproduction.

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