# Image Classification on 15 datasets Average (test)

Source: [Image Classification on 15 datasets Average (test)](<https://www.sota2.com/research/sota/image-classification-on-15-datasets-average-test>)

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

Data updated: 2026-04-23T07:35:04.676Z

Average performance of Image Classification models evaluated across a collection of 15 distinct datasets, specifically reporting Robust Accuracy metrics across varying levels of input perturbation (1/255 to 4/255).

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) | 15 datasets Average (test) |
| Metric | Average Robust Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 17 of 17 |
| 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

- [Average Robust Accuracy](<https://www.sota2.com/research/sota/image-classification-on-15-datasets-average-test.md?metric=Average%20Robust%20Accuracy>): Higher is better; unit: PERCENT.

- [Robust Accuracy (1/255)](<https://www.sota2.com/research/sota/image-classification-on-15-datasets-average-test.md?metric=Robust%20Accuracy%20(1%2F255)>): Higher is better; unit: PERCENT.

- [Robust Accuracy (2/255)](<https://www.sota2.com/research/sota/image-classification-on-15-datasets-average-test.md?metric=Robust%20Accuracy%20(2%2F255)>): Higher is better; unit: PERCENT.

- [Robust Accuracy (3/255)](<https://www.sota2.com/research/sota/image-classification-on-15-datasets-average-test.md?metric=Robust%20Accuracy%20(3%2F255)>): Higher is better; unit: PERCENT.

- [Robust Accuracy (4/255)](<https://www.sota2.com/research/sota/image-classification-on-15-datasets-average-test.md?metric=Robust%20Accuracy%20(4%2F255)>): Higher is better; unit: PERCENT.

- [Robust Accuracy (BERT-Attack)](<https://www.sota2.com/research/sota/image-classification-on-15-datasets-average-test.md?metric=Robust%20Accuracy%20(BERT-Attack)>): Higher is better; unit: PERCENT.

- [Robust Accuracy (GBDA)](<https://www.sota2.com/research/sota/image-classification-on-15-datasets-average-test.md?metric=Robust%20Accuracy%20(GBDA)>): Higher is better; unit: PERCENT.

- [Robust Accuracy (Co-Attack)](<https://www.sota2.com/research/sota/image-classification-on-15-datasets-average-test.md?metric=Robust%20Accuracy%20(Co-Attack)>): Higher is better; unit: PERCENT.

- [Robust Accuracy (SGA)](<https://www.sota2.com/research/sota/image-classification-on-15-datasets-average-test.md?metric=Robust%20Accuracy%20(SGA)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Average Robust Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| Ours (5 trees) | 45.39 | Not recorded | 69e85bd47d0172140ec81123 |
| Ours | 44.34 | Not recorded | 69e85bd47d0172140ec81123 |
| AoS | 43.88 | Not recorded | 69e85bd47d0172140ec81123 |
| PMG-FT | 39.72 | Not recorded | 69e85bd47d0172140ec81123 |
| TeCoA | 38.91 | Not recorded | 69e85bd47d0172140ec81123 |
| FARE | 37.93 | Not recorded | 69e85bd47d0172140ec81123 |
| AGFT | 25.46 | Not recorded | 69cca6c30636e5bf4afcbd96 |
| TGA-ZSR | 23.47 | Not recorded | 69cca6c30636e5bf4afcbd96 |
| GLADIATOR | 23.43 | Not recorded | 69cca6c30636e5bf4afcbd96 |
| PMG-AFT | 21.9 | Not recorded | 69cca6c30636e5bf4afcbd96 |
| TeCoA | 19.86 | Not recorded | 69cca6c30636e5bf4afcbd96 |
| CLIP | 7.21 | Not recorded | 69e85bd47d0172140ec81123 |

### Ours (5 trees)

Average Robust Accuracy: 45.39. Result ID: 69e9cb8a47f34ce924d09e72.

[Reporting paper](<https://www.sota2.com/research/paper/69e85bd47d0172140ec81123.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69e85bd47d0172140ec81123.md>)

Recorded evaluation configuration: {"Zero-shot":"true","Adversarial fine-tuning":"ImageNet","PGD attack steps":"20","Perturbation radius (epsilon)":"1/255","Number of trees":"5"}. This record does not assert independent reproduction.

### Ours

Average Robust Accuracy: 44.34. Result ID: 69e9cb8a47f34ce924d09e71.

[Reporting paper](<https://www.sota2.com/research/paper/69e85bd47d0172140ec81123.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69e85bd47d0172140ec81123.md>)

Recorded evaluation configuration: {"Zero-shot":"true","Adversarial fine-tuning":"ImageNet","PGD attack steps":"20","Perturbation radius (epsilon)":"1/255"}. This record does not assert independent reproduction.

### AoS

Average Robust Accuracy: 43.88. Result ID: 69e9cb8a47f34ce924d09e70.

[Reporting paper](<https://www.sota2.com/research/paper/69e85bd47d0172140ec81123.md>)

Recorded evaluation configuration: {"Zero-shot":"true","Adversarial fine-tuning":"ImageNet","PGD attack steps":"20","Perturbation radius (epsilon)":"1/255"}. This record does not assert independent reproduction.

### PMG-FT

Average Robust Accuracy: 39.72. Result ID: 69e9cb8a47f34ce924d09e6e.

[Reporting paper](<https://www.sota2.com/research/paper/69e85bd47d0172140ec81123.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72df1e209131cea5912e.md>)

Recorded evaluation configuration: {"Zero-shot":"true","Adversarial fine-tuning":"ImageNet","PGD attack steps":"20","Perturbation radius (epsilon)":"1/255"}. This record does not assert independent reproduction.

### TeCoA

Average Robust Accuracy: 38.91. Result ID: 69e9cb8a47f34ce924d09e6d.

[Reporting paper](<https://www.sota2.com/research/paper/69e85bd47d0172140ec81123.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f271e209131cea44177.md>)

Recorded evaluation configuration: {"Zero-shot":"true","Adversarial fine-tuning":"ImageNet","PGD attack steps":"20","Perturbation radius (epsilon)":"1/255"}. This record does not assert independent reproduction.

### FARE

Average Robust Accuracy: 37.93. Result ID: 69e9cb8a47f34ce924d09e6f.

[Reporting paper](<https://www.sota2.com/research/paper/69e85bd47d0172140ec81123.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa738c1e209131cea5d24a.md>)

Recorded evaluation configuration: {"Zero-shot":"true","Adversarial fine-tuning":"ImageNet","PGD attack steps":"20","Perturbation radius (epsilon)":"1/255"}. This record does not assert independent reproduction.

### AGFT

Average Robust Accuracy: 25.46. Result ID: 69ccd3dc47f34ce924b66a85.

[Reporting paper](<https://www.sota2.com/research/paper/69cca6c30636e5bf4afcbd96.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69cca6c30636e5bf4afcbd96.md>)

Recorded evaluation configuration: {"Mode":"Zero-shot","Attack":"PGD-20"}. This record does not assert independent reproduction.

### TGA-ZSR

Average Robust Accuracy: 23.47. Result ID: 69ccd3dc47f34ce924b66a83.

[Reporting paper](<https://www.sota2.com/research/paper/69cca6c30636e5bf4afcbd96.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75771e209131cea68256.md>)

[Implementation](<https://github.com/zhyblue424/TGA-ZSR>)

Recorded evaluation configuration: {"Mode":"Zero-shot","Attack":"PGD-20"}. This record does not assert independent reproduction.

### GLADIATOR

Average Robust Accuracy: 23.43. Result ID: 69ccd3dc47f34ce924b66a84.

[Reporting paper](<https://www.sota2.com/research/paper/69cca6c30636e5bf4afcbd96.md>)

Recorded evaluation configuration: {"Mode":"Zero-shot","Attack":"PGD-20"}. This record does not assert independent reproduction.

### PMG-AFT

Average Robust Accuracy: 21.9. Result ID: 69ccd3dc47f34ce924b66a82.

[Reporting paper](<https://www.sota2.com/research/paper/69cca6c30636e5bf4afcbd96.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72df1e209131cea5912e.md>)

Recorded evaluation configuration: {"Mode":"Zero-shot","Attack":"PGD-20"}. This record does not assert independent reproduction.

### TeCoA

Average Robust Accuracy: 19.86. Result ID: 69ccd3dc47f34ce924b66a81.

[Reporting paper](<https://www.sota2.com/research/paper/69cca6c30636e5bf4afcbd96.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f271e209131cea44177.md>)

Recorded evaluation configuration: {"Mode":"Zero-shot","Attack":"PGD-20"}. This record does not assert independent reproduction.

### CLIP

Average Robust Accuracy: 7.21. Result ID: 69e9cb8a47f34ce924d09e6c.

[Reporting paper](<https://www.sota2.com/research/paper/69e85bd47d0172140ec81123.md>)

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

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

Recorded evaluation configuration: {"Zero-shot":"true","Adversarial fine-tuning":"ImageNet","PGD attack steps":"20","Perturbation radius (epsilon)":"1/255"}. This record does not assert independent reproduction.

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