# Image Classification on ImageNet Distribution Shifts IN-R IN-A IN-S V2

Source: [Image Classification on ImageNet Distribution Shifts IN-R IN-A IN-S V2](<https://www.sota2.com/research/sota/image-classification-on-imagenet-distribution-shifts-in-r-in-a-in-s-v2>)

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

Data updated: 2026-05-18T08:49:30.757Z

Evaluation of Image Classification performance across the core ImageNet dataset and its distribution shift variants: ImageNet-R, ImageNet-A, ImageNet-S (Sketch), and ImageNet-V2.

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 | Image Classification |
| Dataset (including any stated split) | ImageNet Distribution Shifts IN-R IN-A IN-S V2 |
| Metric | ImageNet Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 4 of 4 |
| 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

- [ImageNet Accuracy](<https://www.sota2.com/research/sota/image-classification-on-imagenet-distribution-shifts-in-r-in-a-in-s-v2.md?metric=ImageNet%20Accuracy>): Higher is better; unit: PERCENT.

- [ImageNet-R Accuracy](<https://www.sota2.com/research/sota/image-classification-on-imagenet-distribution-shifts-in-r-in-a-in-s-v2.md?metric=ImageNet-R%20Accuracy>): Higher is better; unit: PERCENT.

- [ImageNet-A Accuracy](<https://www.sota2.com/research/sota/image-classification-on-imagenet-distribution-shifts-in-r-in-a-in-s-v2.md?metric=ImageNet-A%20Accuracy>): Higher is better; unit: PERCENT.

- [ImageNet-S Accuracy](<https://www.sota2.com/research/sota/image-classification-on-imagenet-distribution-shifts-in-r-in-a-in-s-v2.md?metric=ImageNet-S%20Accuracy>): Higher is better; unit: PERCENT.

- [ImageNet-V2 Accuracy](<https://www.sota2.com/research/sota/image-classification-on-imagenet-distribution-shifts-in-r-in-a-in-s-v2.md?metric=ImageNet-V2%20Accuracy>): Higher is better; unit: PERCENT.

- [Average Accuracy](<https://www.sota2.com/research/sota/image-classification-on-imagenet-distribution-shifts-in-r-in-a-in-s-v2.md?metric=Average%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | ImageNet Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| SAE Add | 85.3 | Not recorded | 6a0a9d2cdbc836b87d82b4d7 |
| WiSE-FT | 84.8 | Not recorded | 6a0a9d2cdbc836b87d82b4d7 |
| FT | 82.8 | Not recorded | 6a0a9d2cdbc836b87d82b4d7 |
| Zero-shot | 78.5 | Not recorded | 6a0a9d2cdbc836b87d82b4d7 |

### SAE Add

ImageNet Accuracy: 85.3. Result ID: 6a0ad28e47f34ce924f11763.

[Reporting paper](<https://www.sota2.com/research/paper/6a0a9d2cdbc836b87d82b4d7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a0a9d2cdbc836b87d82b4d7.md>)

[Implementation](<https://github.com/Fabian-Mor/sae-ft>)

Recorded evaluation configuration: {"Backbone":"SigLIP2 ViT-B/16"}. This record does not assert independent reproduction.

### WiSE-FT

ImageNet Accuracy: 84.8. Result ID: 6a0ad28e47f34ce924f11762.

[Reporting paper](<https://www.sota2.com/research/paper/6a0a9d2cdbc836b87d82b4d7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6cd31e209131cea3601a.md>)

[Implementation](<https://github.com/mlfoundations/wise-ft>)

Recorded evaluation configuration: {"Backbone":"SigLIP2 ViT-B/16"}. This record does not assert independent reproduction.

### FT

ImageNet Accuracy: 82.8. Result ID: 6a0ad28e47f34ce924f11761.

[Reporting paper](<https://www.sota2.com/research/paper/6a0a9d2cdbc836b87d82b4d7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79af1e209131cea7edf0.md>)

Recorded evaluation configuration: {"Backbone":"SigLIP2 ViT-B/16"}. This record does not assert independent reproduction.

### Zero-shot

ImageNet Accuracy: 78.5. Result ID: 6a0ad28e47f34ce924f11760.

[Reporting paper](<https://www.sota2.com/research/paper/6a0a9d2cdbc836b87d82b4d7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69581e209131cea21ad0.md>)

[Implementation](<https://github.com/OpenAI/CLIP>)

Recorded evaluation configuration: {"Backbone":"SigLIP2 ViT-B/16"}. This record does not assert independent reproduction.

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