# Image Classification on ImageNet (val) (Epsilon-based Top-1 Accuracy)

Source: [Image Classification on ImageNet (val) (Epsilon-based Top-1 Accuracy)](<https://www.sota2.com/research/sota/image-classification-on-imagenet-val-epsilon-based-top-1-accuracy>)

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

Data updated: 2026-02-26T06:06:54.476Z

Evaluation of Image Classification performance on the validation split of the ImageNet dataset, reporting Top-1 Accuracy measured under various epsilon conditions (1, 2, 4, 8).

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 (val) |
| Metric | Top-1 Acc (eps=8) |
| 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

- [Top-1 Acc (eps=8)](<https://www.sota2.com/research/sota/image-classification-on-imagenet-val-epsilon-based-top-1-accuracy.md?metric=Top-1%20Acc%20(eps%3D8)>): Higher is better; unit: PERCENT.

- [Top-1 Acc (eps=1)](<https://www.sota2.com/research/sota/image-classification-on-imagenet-val-epsilon-based-top-1-accuracy.md?metric=Top-1%20Acc%20(eps%3D1)>): Higher is better; unit: PERCENT.

- [Top-1 Acc (eps=2)](<https://www.sota2.com/research/sota/image-classification-on-imagenet-val-epsilon-based-top-1-accuracy.md?metric=Top-1%20Acc%20(eps%3D2)>): Higher is better; unit: PERCENT.

- [Top-1 Acc (eps=4)](<https://www.sota2.com/research/sota/image-classification-on-imagenet-val-epsilon-based-top-1-accuracy.md?metric=Top-1%20Acc%20(eps%3D4)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Top-1 Acc (eps=8) | Result date | Report paper |
| --- | --- | --- | --- |
| DP-SGD | 86.7 | Not recorded | 68fa6cbf1e209131cea35c41 |
| DP-SGD | 32.4 | Not recorded | 68fa6cbf1e209131cea35c41 |

### DP-SGD

Top-1 Acc (eps=8): 86.7. Result ID: 698f3afb3a6c314849f74cff.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6cbf1e209131cea35c41.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6cbf1e209131cea35c41.md>)

[Implementation](<https://github.com/deepmind/jax_privacy>)

Recorded evaluation configuration: {"Pre-Training Data":"JFT-4B","Architecture":"NFNet-F3","delta":"8 · 10^-7"}. This record does not assert independent reproduction.

### DP-SGD

Top-1 Acc (eps=8): 32.4. Result ID: 698f3afb3a6c314849f74cfe.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6cbf1e209131cea35c41.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6cbf1e209131cea35c41.md>)

[Implementation](<https://github.com/deepmind/jax_privacy>)

Recorded evaluation configuration: {"Pre-Training Data":"None","Architecture":"NF-ResNet-50","delta":"8 · 10^-7"}. This record does not assert independent reproduction.

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