# Metric correlation with human assessment of privacy leakage on CelebA (test)

Source: [Metric correlation with human assessment of privacy leakage on CelebA (test)](<https://www.sota2.com/research/sota/metric-correlation-with-human-assessment-of-privacy-leakage-on-celeba-test>)

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

Data updated: 2026-02-26T06:04:38.186Z

Evaluation of model performance based on the correlation between predicted privacy leakage metrics and human assessments, using the test split of the CelebA dataset. Metrics reported include Spearman's rho and Kendall's tau.

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 | Metric correlation with human assessment of privacy leakage |
| Dataset (including any stated split) | CelebA (test) |
| Metric | Spearman's Rho |
| Unit | SCALAR |
| Direction | Higher 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

- [Spearman's Rho](<https://www.sota2.com/research/sota/metric-correlation-with-human-assessment-of-privacy-leakage-on-celeba-test.md?metric=Spearman%27s%20Rho>): Higher is better; unit: SCALAR.

- [Kendall's Tau](<https://www.sota2.com/research/sota/metric-correlation-with-human-assessment-of-privacy-leakage-on-celeba-test.md?metric=Kendall%27s%20Tau>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Spearman's Rho | Result date | Report paper |
| --- | --- | --- | --- |
| PSNR | 0.7495 | Not recorded | 68fa70c91e209131cea4d1ad |
| SSIM | 0.6846 | Not recorded | 68fa70c91e209131cea4d1ad |
| FID | -0.1516 | Not recorded | 68fa70c91e209131cea4d1ad |
| LPIPS | -0.5824 | Not recorded | 68fa70c91e209131cea4d1ad |
| MSE | -0.7349 | Not recorded | 68fa70c91e209131cea4d1ad |
| SemSim | -0.8263 | Not recorded | 68fa70c91e209131cea4d1ad |

### PSNR

Spearman's Rho: 0.7495. Result ID: 6991417e3a6c31484906d38d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa70c91e209131cea4d1ad.md>)

Recorded evaluation configuration: {"Attack method":"InvGrad \[7\]","Evaluation basis":"Ranking of 14 target models"}. This record does not assert independent reproduction.

### SSIM

Spearman's Rho: 0.6846. Result ID: 6991417e3a6c31484906d38f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa70c91e209131cea4d1ad.md>)

Recorded evaluation configuration: {"Attack method":"InvGrad \[7\]","Evaluation basis":"Ranking of 14 target models"}. This record does not assert independent reproduction.

### FID

Spearman's Rho: -0.1516. Result ID: 6991417e3a6c31484906d391.

[Reporting paper](<https://www.sota2.com/research/paper/68fa70c91e209131cea4d1ad.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa64081e209131cea02279.md>)

[Implementation](<https://github.com/bioinf-jku/TTUR>)

Recorded evaluation configuration: {"Attack method":"InvGrad \[7\]","Evaluation basis":"Ranking of 14 target models"}. This record does not assert independent reproduction.

### LPIPS

Spearman's Rho: -0.5824. Result ID: 6991417e3a6c31484906d390.

[Reporting paper](<https://www.sota2.com/research/paper/68fa70c91e209131cea4d1ad.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa64441e209131cea039ef.md>)

Recorded evaluation configuration: {"Attack method":"InvGrad \[7\]","Evaluation basis":"Ranking of 14 target models"}. This record does not assert independent reproduction.

### MSE

Spearman's Rho: -0.7349. Result ID: 6991417e3a6c31484906d38e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa70c91e209131cea4d1ad.md>)

Recorded evaluation configuration: {"Attack method":"InvGrad \[7\]","Evaluation basis":"Ranking of 14 target models"}. This record does not assert independent reproduction.

### SemSim

Spearman's Rho: -0.8263. Result ID: 6991417e3a6c31484906d392.

[Reporting paper](<https://www.sota2.com/research/paper/68fa70c91e209131cea4d1ad.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70c91e209131cea4d1ad.md>)

[Implementation](<https://sites.google.com/view/semsim>)

Recorded evaluation configuration: {"Attack method":"InvGrad \[7\]","Evaluation basis":"Ranking of 14 target models","Training setup":"Leave-one-out on 4 other datasets"}. This record does not assert independent reproduction.

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