# Face Verification on RFW (test) (Calibration and Group Fairness)

Source: [Face Verification on RFW (test) (Calibration and Group Fairness)](<https://www.sota2.com/research/sota/face-verification-on-rfw-test-calibration-and-group-fairness>)

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

Data updated: 2026-06-04T12:09:20.065Z

Evaluation of face verification performance on the test split of the Racial Face in the Wild (RFW) dataset, focusing on calibration metrics (ECE, Brier) and group performance disparities (TPR, PE Gap, Min-grp AUROC).

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 | Face Verification |
| Dataset (including any stated split) | RFW (test) |
| Metric | AUROC (%) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 33 of 33 |
| 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

- [AUROC (%)](<https://www.sota2.com/research/sota/face-verification-on-rfw-test-calibration-and-group-fairness.md?metric=AUROC%20(%25)>): Higher is better; unit: PERCENT.

- [TPR (0.1%)](<https://www.sota2.com/research/sota/face-verification-on-rfw-test-calibration-and-group-fairness.md?metric=TPR%20(0.1%25)>): Higher is better; unit: PERCENT.

- [TPR (1.0%)](<https://www.sota2.com/research/sota/face-verification-on-rfw-test-calibration-and-group-fairness.md?metric=TPR%20(1.0%25)>): Higher is better; unit: PERCENT.

- [PE Gap (0.1%)](<https://www.sota2.com/research/sota/face-verification-on-rfw-test-calibration-and-group-fairness.md?metric=PE%20Gap%20(0.1%25)>): Lower is better; unit: PERCENT.

- [PE Gap (1.0%)](<https://www.sota2.com/research/sota/face-verification-on-rfw-test-calibration-and-group-fairness.md?metric=PE%20Gap%20(1.0%25)>): Lower is better; unit: PERCENT.

- [ECE](<https://www.sota2.com/research/sota/face-verification-on-rfw-test-calibration-and-group-fairness.md?metric=ECE>): Lower is better; unit: PERCENT.

- [Brier](<https://www.sota2.com/research/sota/face-verification-on-rfw-test-calibration-and-group-fairness.md?metric=Brier>): Lower is better; unit: PERCENT.

- [Min-grp AUROC (%)](<https://www.sota2.com/research/sota/face-verification-on-rfw-test-calibration-and-group-fairness.md?metric=Min-grp%20AUROC%20(%25)>): Higher is better; unit: PERCENT.

- [LD-TPR@0.1](<https://www.sota2.com/research/sota/face-verification-on-rfw-test-calibration-and-group-fairness.md?metric=LD-TPR%400.1>): Higher is better; unit: PERCENT.

## Recorded results

| Method | AUROC (%) | Result date | Report paper |
| --- | --- | --- | --- |
| AC-Ridge | 99.71 | Not recorded | 6a2106f9c10484bac23f06bd |
| AC-MLP | 99.33 | Not recorded | 6a2106f9c10484bac23f06bd |
| AC-Linear | 99.31 | Not recorded | 6a2106f9c10484bac23f06bd |
| FairCal | 99.26 | Not recorded | 6a2106f9c10484bac23f06bd |
| Oracle† | 99.26 | Not recorded | 6a2106f9c10484bac23f06bd |
| Baseline | 99.25 | Not recorded | 6a2106f9c10484bac23f06bd |
| FALCON | 99.25 | Not recorded | 6a2106f9c10484bac23f06bd |
| GST† | 99.24 | Not recorded | 6a2106f9c10484bac23f06bd |
| DemoNorm† | 99.24 | Not recorded | 6a2106f9c10484bac23f06bd |
| FRAPPE | 99.2 | Not recorded | 6a2106f9c10484bac23f06bd |
| FSN | 99.13 | Not recorded | 6a2106f9c10484bac23f06bd |
| AC-Ridge | 98.95 | Not recorded | 6a2106f9c10484bac23f06bd |
| AC-Linear | 98.74 | Not recorded | 6a2106f9c10484bac23f06bd |
| AC-MLP | 98.66 | Not recorded | 6a2106f9c10484bac23f06bd |
| FairCal | 98.63 | Not recorded | 6a2106f9c10484bac23f06bd |
| Oracle | 98.52 | Not recorded | 6a2106f9c10484bac23f06bd |
| DemoNorm | 98.46 | Not recorded | 6a2106f9c10484bac23f06bd |
| FRAPPÉ | 98.43 | Not recorded | 6a2106f9c10484bac23f06bd |
| AC-Ridge | 98.43 | Not recorded | 6a2106f9c10484bac23f06bd |
| FALCON | 98.41 | Not recorded | 6a2106f9c10484bac23f06bd |
| GST | 98.4 | Not recorded | 6a2106f9c10484bac23f06bd |
| Baseline | 98.39 | Not recorded | 6a2106f9c10484bac23f06bd |
| FSN | 98.12 | Not recorded | 6a2106f9c10484bac23f06bd |
| AC-Linear | 98.04 | Not recorded | 6a2106f9c10484bac23f06bd |
| AC-MLP | 97.98 | Not recorded | 6a2106f9c10484bac23f06bd |
| FairCal | 97.83 | Not recorded | 6a2106f9c10484bac23f06bd |
| Oracle | 97.58 | Not recorded | 6a2106f9c10484bac23f06bd |
| DemoNorm | 97.48 | Not recorded | 6a2106f9c10484bac23f06bd |
| FALCON | 97.46 | Not recorded | 6a2106f9c10484bac23f06bd |
| Baseline | 97.42 | Not recorded | 6a2106f9c10484bac23f06bd |
| GST | 97.42 | Not recorded | 6a2106f9c10484bac23f06bd |
| FRAPPE | 97.33 | Not recorded | 6a2106f9c10484bac23f06bd |
| FSN | 97.16 | Not recorded | 6a2106f9c10484bac23f06bd |

### AC-Ridge

AUROC (%): 99.71. Result ID: 6a216ad6fbc373592c1e6972.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

Recorded evaluation configuration: {"Backbone":"FaceNet","Demographic labels at inference":false}. This record does not assert independent reproduction.

### AC-MLP

AUROC (%): 99.33. Result ID: 6a216ad6fbc373592c1e6970.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

Recorded evaluation configuration: {"Backbone":"FaceNet","Demographic labels at inference":false}. This record does not assert independent reproduction.

### AC-Linear

AUROC (%): 99.31. Result ID: 6a216ad6fbc373592c1e6971.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

Recorded evaluation configuration: {"Backbone":"FaceNet","Demographic labels at inference":false}. This record does not assert independent reproduction.

### FairCal

AUROC (%): 99.26. Result ID: 6a216ad6fbc373592c1e6969.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c341e209131cea32cb0.md>)

Recorded evaluation configuration: {"Backbone":"FaceNet","Demographic labels at inference":false}. This record does not assert independent reproduction.

### Oracle†

AUROC (%): 99.26. Result ID: 6a216ad6fbc373592c1e696d.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c341e209131cea32cb0.md>)

Recorded evaluation configuration: {"Backbone":"FaceNet","Demographic labels at inference":true}. This record does not assert independent reproduction.

### Baseline

AUROC (%): 99.25. Result ID: 6a216ad6fbc373592c1e6968.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

Recorded evaluation configuration: {"Backbone":"FaceNet","Demographic labels at inference":false}. This record does not assert independent reproduction.

### FALCON

AUROC (%): 99.25. Result ID: 6a216ad6fbc373592c1e696a.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

Recorded evaluation configuration: {"Backbone":"FaceNet","Demographic labels at inference":false}. This record does not assert independent reproduction.

### GST†

AUROC (%): 99.24. Result ID: 6a216ad6fbc373592c1e696e.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa676e1e209131cea1624b.md>)

Recorded evaluation configuration: {"Backbone":"FaceNet","Demographic labels at inference":true}. This record does not assert independent reproduction.

### DemoNorm†

AUROC (%): 99.24. Result ID: 6a216ad6fbc373592c1e696f.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa742c1e209131cea602f9.md>)

Recorded evaluation configuration: {"Backbone":"FaceNet","Demographic labels at inference":true}. This record does not assert independent reproduction.

### FRAPPE

AUROC (%): 99.2. Result ID: 6a216ad6fbc373592c1e696c.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73c31e209131cea5e50d.md>)

Recorded evaluation configuration: {"Backbone":"FaceNet","Demographic labels at inference":false}. This record does not assert independent reproduction.

### FSN

AUROC (%): 99.13. Result ID: 6a216ad6fbc373592c1e696b.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa68a61e209131cea1d6a4.md>)

Recorded evaluation configuration: {"Backbone":"FaceNet","Demographic labels at inference":false}. This record does not assert independent reproduction.

### AC-Ridge

AUROC (%): 98.95. Result ID: 6a216ad6fbc373592c1e6730.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

Recorded evaluation configuration: {"Backbone":"MagFace","Protocol":"Label-free"}. This record does not assert independent reproduction.

### AC-Linear

AUROC (%): 98.74. Result ID: 6a216ad6fbc373592c1e672f.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

Recorded evaluation configuration: {"Backbone":"MagFace","Protocol":"Label-free"}. This record does not assert independent reproduction.

### AC-MLP

AUROC (%): 98.66. Result ID: 6a216ad6fbc373592c1e672e.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

Recorded evaluation configuration: {"Backbone":"MagFace","Protocol":"Label-free"}. This record does not assert independent reproduction.

### FairCal

AUROC (%): 98.63. Result ID: 6a216ad6fbc373592c1e6727.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c341e209131cea32cb0.md>)

Recorded evaluation configuration: {"Backbone":"MagFace","Protocol":"Label-free"}. This record does not assert independent reproduction.

### Oracle

AUROC (%): 98.52. Result ID: 6a216ad6fbc373592c1e672b.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c341e209131cea32cb0.md>)

Recorded evaluation configuration: {"Backbone":"MagFace","Protocol":"Demographic-aware"}. This record does not assert independent reproduction.

### DemoNorm

AUROC (%): 98.46. Result ID: 6a216ad6fbc373592c1e672d.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa742c1e209131cea602f9.md>)

Recorded evaluation configuration: {"Backbone":"MagFace","Protocol":"Demographic-aware"}. This record does not assert independent reproduction.

### FRAPPÉ

AUROC (%): 98.43. Result ID: 6a216ad6fbc373592c1e672a.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73c31e209131cea5e50d.md>)

Recorded evaluation configuration: {"Backbone":"MagFace","Protocol":"Label-free"}. This record does not assert independent reproduction.

### AC-Ridge

AUROC (%): 98.43. Result ID: 6a216ad6fbc373592c1e67c9.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

Recorded evaluation configuration: {"Backbone":"ArcFace R100","Demographic-aware":"false"}. This record does not assert independent reproduction.

### FALCON

AUROC (%): 98.41. Result ID: 6a216ad6fbc373592c1e6728.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

Recorded evaluation configuration: {"Backbone":"MagFace","Protocol":"Label-free"}. This record does not assert independent reproduction.

### GST

AUROC (%): 98.4. Result ID: 6a216ad6fbc373592c1e672c.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa676e1e209131cea1624b.md>)

Recorded evaluation configuration: {"Backbone":"MagFace","Protocol":"Demographic-aware"}. This record does not assert independent reproduction.

### Baseline

AUROC (%): 98.39. Result ID: 6a216ad6fbc373592c1e6726.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

Recorded evaluation configuration: {"Backbone":"MagFace","Protocol":"Label-free"}. This record does not assert independent reproduction.

### FSN

AUROC (%): 98.12. Result ID: 6a216ad6fbc373592c1e6729.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa68a61e209131cea1d6a4.md>)

Recorded evaluation configuration: {"Backbone":"MagFace","Protocol":"Label-free"}. This record does not assert independent reproduction.

### AC-Linear

AUROC (%): 98.04. Result ID: 6a216ad6fbc373592c1e67c8.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

Recorded evaluation configuration: {"Backbone":"ArcFace R100","Demographic-aware":"false"}. This record does not assert independent reproduction.

### AC-MLP

AUROC (%): 97.98. Result ID: 6a216ad6fbc373592c1e67c7.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

Recorded evaluation configuration: {"Backbone":"ArcFace R100","Demographic-aware":"false"}. This record does not assert independent reproduction.

### FairCal

AUROC (%): 97.83. Result ID: 6a216ad6fbc373592c1e67c0.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c341e209131cea32cb0.md>)

Recorded evaluation configuration: {"Backbone":"ArcFace R100","Demographic-aware":"false"}. This record does not assert independent reproduction.

### Oracle

AUROC (%): 97.58. Result ID: 6a216ad6fbc373592c1e67c4.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c341e209131cea32cb0.md>)

Recorded evaluation configuration: {"Backbone":"ArcFace R100","Demographic-aware":"true"}. This record does not assert independent reproduction.

### DemoNorm

AUROC (%): 97.48. Result ID: 6a216ad6fbc373592c1e67c6.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa742c1e209131cea602f9.md>)

Recorded evaluation configuration: {"Backbone":"ArcFace R100","Demographic-aware":"true"}. This record does not assert independent reproduction.

### FALCON

AUROC (%): 97.46. Result ID: 6a216ad6fbc373592c1e67c1.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

Recorded evaluation configuration: {"Backbone":"ArcFace R100","Demographic-aware":"false"}. This record does not assert independent reproduction.

### Baseline

AUROC (%): 97.42. Result ID: 6a216ad6fbc373592c1e67bf.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

Recorded evaluation configuration: {"Backbone":"ArcFace R100","Demographic-aware":"false"}. This record does not assert independent reproduction.

### GST

AUROC (%): 97.42. Result ID: 6a216ad6fbc373592c1e67c5.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa676e1e209131cea1624b.md>)

Recorded evaluation configuration: {"Backbone":"ArcFace R100","Demographic-aware":"true"}. This record does not assert independent reproduction.

### FRAPPE

AUROC (%): 97.33. Result ID: 6a216ad6fbc373592c1e67c3.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73c31e209131cea5e50d.md>)

Recorded evaluation configuration: {"Backbone":"ArcFace R100","Demographic-aware":"false"}. This record does not assert independent reproduction.

### FSN

AUROC (%): 97.16. Result ID: 6a216ad6fbc373592c1e67c2.

[Reporting paper](<https://www.sota2.com/research/paper/6a2106f9c10484bac23f06bd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa68a61e209131cea1d6a4.md>)

Recorded evaluation configuration: {"Backbone":"ArcFace R100","Demographic-aware":"false"}. This record does not assert independent reproduction.

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