# Image Perturbation Quality Assessment on ImageNet Human Evaluation

Source: [Image Perturbation Quality Assessment on ImageNet Human Evaluation](<https://www.sota2.com/research/sota/image-perturbation-quality-assessment-on-imagenet-human-evaluation>)

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

Data updated: 2026-02-26T05:34:53.881Z

Human evaluation study assessing the quality of perturbed images on the ImageNet dataset, focusing on label preservation, image realism, and ambiguity.

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 Perturbation Quality Assessment |
| Dataset (including any stated split) | ImageNet Human Evaluation |
| Metric | Ambiguity Cases |
| Unit | PERCENT |
| Direction | Lower is better |
| Results loaded | 3 of 3 |
| 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

- [Ambiguity Cases](<https://www.sota2.com/research/sota/image-perturbation-quality-assessment-on-imagenet-human-evaluation.md?metric=Ambiguity%20Cases>): Lower is better; unit: PERCENT.

- [Label Preservation](<https://www.sota2.com/research/sota/image-perturbation-quality-assessment-on-imagenet-human-evaluation.md?metric=Label%20Preservation>): Higher is better; unit: SCALAR.

- [Image Realism](<https://www.sota2.com/research/sota/image-perturbation-quality-assessment-on-imagenet-human-evaluation.md?metric=Image%20Realism>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Ambiguity Cases | Result date | Report paper |
| --- | --- | --- | --- |
| HyNeA | 2.5 | Not recorded | 6971af33e25ad2fa9d1edb51 |
| Mimicry | 7.5 | Not recorded | 6971af33e25ad2fa9d1edb51 |
| GIFTbench | 7.5 | Not recorded | 6971af33e25ad2fa9d1edb51 |

### HyNeA

Ambiguity Cases: 2.5. Result ID: 699925a15698665ec48d2373.

[Reporting paper](<https://www.sota2.com/research/paper/6971af33e25ad2fa9d1edb51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6971af33e25ad2fa9d1edb51.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### Mimicry

Ambiguity Cases: 7.5. Result ID: 699925a15698665ec48d2374.

[Reporting paper](<https://www.sota2.com/research/paper/6971af33e25ad2fa9d1edb51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78741e209131cea7866e.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### GIFTbench

Ambiguity Cases: 7.5. Result ID: 699925a15698665ec48d2375.

[Reporting paper](<https://www.sota2.com/research/paper/6971af33e25ad2fa9d1edb51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa765f1e209131cea6cb21.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

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