# Racial Bias Evaluation on Multilingual CrowS-Pairs

Source: [Racial Bias Evaluation on Multilingual CrowS-Pairs](<https://www.sota2.com/research/sota/racial-bias-evaluation-on-multilingual-crows-pairs>)

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

Data updated: 2026-04-06T08:14:26.671Z

Evaluation of racial bias performance across multiple languages (German, Chinese, Spanish, Japanese) using the Multilingual CrowS-Pairs dataset.

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 | Racial Bias Evaluation |
| Dataset (including any stated split) | Multilingual CrowS-Pairs racial bias |
| Metric | Bias Score (DE) |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 18 of 18 |
| 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

- [Bias Score (DE)](<https://www.sota2.com/research/sota/racial-bias-evaluation-on-multilingual-crows-pairs.md?metric=Bias%20Score%20(DE)>): Higher is better; unit: SCALAR.

- [Bias Score (ZH)](<https://www.sota2.com/research/sota/racial-bias-evaluation-on-multilingual-crows-pairs.md?metric=Bias%20Score%20(ZH)>): Higher is better; unit: SCALAR.

- [Bias Score (ES)](<https://www.sota2.com/research/sota/racial-bias-evaluation-on-multilingual-crows-pairs.md?metric=Bias%20Score%20(ES)>): Higher is better; unit: SCALAR.

- [Bias Score (JA)](<https://www.sota2.com/research/sota/racial-bias-evaluation-on-multilingual-crows-pairs.md?metric=Bias%20Score%20(JA)>): Higher is better; unit: SCALAR.

- [Average Bias Score](<https://www.sota2.com/research/sota/racial-bias-evaluation-on-multilingual-crows-pairs.md?metric=Average%20Bias%20Score>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Bias Score (DE) | Result date | Report paper |
| --- | --- | --- | --- |
| XLM-R | 16.37 | Not recorded | 69d33dd97d0172140ec734ee |
| CDA | 15.56 | Not recorded | 69d33dd97d0172140ec734ee |
| SD | 15.49 | Not recorded | 69d33dd97d0172140ec734ee |
| Multiple-Debias (MD) | 11.41 | Not recorded | 69d33dd97d0172140ec734ee |
| Multiple-Debias (MD) | 10.56 | Not recorded | 69d33dd97d0172140ec734ee |
| Multiple-Debias (MD) | 10.56 | Not recorded | 69d33dd97d0172140ec734ee |
| MSD | 10.43 | Not recorded | 69d33dd97d0172140ec734ee |
| MCDA | 8.01 | Not recorded | 69d33dd97d0172140ec734ee |
| Multiple-Debias (MD) | 8.01 | Not recorded | 69d33dd97d0172140ec734ee |
| Multiple-Debias (MD) | 7.88 | Not recorded | 69d33dd97d0172140ec734ee |
| SD | 7.24 | Not recorded | 69d33dd97d0172140ec734ee |
| MCDA | 5.71 | Not recorded | 69d33dd97d0172140ec734ee |
| MSD | 5.56 | Not recorded | 69d33dd97d0172140ec734ee |
| mBERT | 4.74 | Not recorded | 69d33dd97d0172140ec734ee |
| Multiple-Debias (MD) | 2.93 | Not recorded | 69d33dd97d0172140ec734ee |
| CDA | 2.01 | Not recorded | 69d33dd97d0172140ec734ee |
| Multiple-Debias (MD) | 0.56 | Not recorded | 69d33dd97d0172140ec734ee |
| Multiple-Debias (MD) | 0.53 | Not recorded | 69d33dd97d0172140ec734ee |

### XLM-R

Bias Score (DE): 16.37. Result ID: 69d36b6047f34ce924bb27ec.

[Reporting paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67521e209131cea15706.md>)

Recorded evaluation configuration: {"Base Model":"XLM-R"}. This record does not assert independent reproduction.

### CDA

Bias Score (DE): 15.56. Result ID: 69d36b6047f34ce924bb27ed.

[Reporting paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa643f1e209131cea03715.md>)

Recorded evaluation configuration: {"Base Model":"XLM-R"}. This record does not assert independent reproduction.

### SD

Bias Score (DE): 15.49. Result ID: 69d36b6047f34ce924bb27ee.

[Reporting paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ab81e209131cea2a074.md>)

Recorded evaluation configuration: {"Base Model":"XLM-R"}. This record does not assert independent reproduction.

### Multiple-Debias (MD)

Bias Score (DE): 11.41. Result ID: 69d36b6047f34ce924bb27e9.

[Reporting paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

Recorded evaluation configuration: {"Base Model":"mBERT","Tuning Strategy":"Adapter Tune"}. This record does not assert independent reproduction.

### Multiple-Debias (MD)

Bias Score (DE): 10.56. Result ID: 69d36b6047f34ce924bb27ea.

[Reporting paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

Recorded evaluation configuration: {"Base Model":"mBERT","Tuning Strategy":"Prompt Tune"}. This record does not assert independent reproduction.

### Multiple-Debias (MD)

Bias Score (DE): 10.56. Result ID: 69d36b6047f34ce924bb27eb.

[Reporting paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

Recorded evaluation configuration: {"Base Model":"mBERT","Tuning Strategy":"Prefix Tune"}. This record does not assert independent reproduction.

### MSD

Bias Score (DE): 10.43. Result ID: 69d36b6047f34ce924bb27f0.

[Reporting paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

Recorded evaluation configuration: {"Base Model":"XLM-R"}. This record does not assert independent reproduction.

### MCDA

Bias Score (DE): 8.01. Result ID: 69d36b6047f34ce924bb27ef.

[Reporting paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

Recorded evaluation configuration: {"Base Model":"XLM-R"}. This record does not assert independent reproduction.

### Multiple-Debias (MD)

Bias Score (DE): 8.01. Result ID: 69d36b6047f34ce924bb27f2.

[Reporting paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

Recorded evaluation configuration: {"Base Model":"XLM-R","Tuning Strategy":"Adapter Tune"}. This record does not assert independent reproduction.

### Multiple-Debias (MD)

Bias Score (DE): 7.88. Result ID: 69d36b6047f34ce924bb27f3.

[Reporting paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

Recorded evaluation configuration: {"Base Model":"XLM-R","Tuning Strategy":"Prompt Tune"}. This record does not assert independent reproduction.

### SD

Bias Score (DE): 7.24. Result ID: 69d36b6047f34ce924bb27e5.

[Reporting paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ab81e209131cea2a074.md>)

Recorded evaluation configuration: {"Base Model":"mBERT"}. This record does not assert independent reproduction.

### MCDA

Bias Score (DE): 5.71. Result ID: 69d36b6047f34ce924bb27e6.

[Reporting paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

Recorded evaluation configuration: {"Base Model":"mBERT"}. This record does not assert independent reproduction.

### MSD

Bias Score (DE): 5.56. Result ID: 69d36b6047f34ce924bb27e7.

[Reporting paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

Recorded evaluation configuration: {"Base Model":"mBERT"}. This record does not assert independent reproduction.

### mBERT

Bias Score (DE): 4.74. Result ID: 69d36b6047f34ce924bb27e3.

[Reporting paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65c31e209131cea0c4ed.md>)

[Implementation](<https://github.com/google-research/bert>)

Recorded evaluation configuration: {"Base Model":"mBERT"}. This record does not assert independent reproduction.

### Multiple-Debias (MD)

Bias Score (DE): 2.93. Result ID: 69d36b6047f34ce924bb27e8.

[Reporting paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

Recorded evaluation configuration: {"Base Model":"mBERT","Tuning Strategy":"Full Fine-Tune"}. This record does not assert independent reproduction.

### CDA

Bias Score (DE): 2.01. Result ID: 69d36b6047f34ce924bb27e4.

[Reporting paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa643f1e209131cea03715.md>)

Recorded evaluation configuration: {"Base Model":"mBERT"}. This record does not assert independent reproduction.

### Multiple-Debias (MD)

Bias Score (DE): 0.56. Result ID: 69d36b6047f34ce924bb27f4.

[Reporting paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

Recorded evaluation configuration: {"Base Model":"XLM-R","Tuning Strategy":"Prefix Tune"}. This record does not assert independent reproduction.

### Multiple-Debias (MD)

Bias Score (DE): 0.53. Result ID: 69d36b6047f34ce924bb27f1.

[Reporting paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d33dd97d0172140ec734ee.md>)

Recorded evaluation configuration: {"Base Model":"XLM-R","Tuning Strategy":"Full Fine-Tune"}. This record does not assert independent reproduction.

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