# Gender Bias in Coreference Resolution on WinoBias

Source: [Gender Bias in Coreference Resolution on WinoBias](<https://www.sota2.com/research/sota/gender-bias-in-coreference-resolution-on-winobias>)

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

Data updated: 2026-02-26T05:54:40.127Z

Evaluation of model performance on the WinoBias dataset, specifically measuring gender bias using stereo/anti-stereotype metrics (P\_stereo, P\_anti, Gap).

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 | Gender Bias in Coreference Resolution |
| Dataset (including any stated split) | WinoBias |
| Metric | P(Stereo) |
| Unit | PERCENT |
| Direction | Lower is better |
| Results loaded | 7 of 7 |
| 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

- [P(Stereo)](<https://www.sota2.com/research/sota/gender-bias-in-coreference-resolution-on-winobias.md?metric=P(Stereo)>): Lower is better; unit: PERCENT.

- [P(Anti-Stereo)](<https://www.sota2.com/research/sota/gender-bias-in-coreference-resolution-on-winobias.md?metric=P(Anti-Stereo)>): Higher is better; unit: PERCENT.

- [P(Other)](<https://www.sota2.com/research/sota/gender-bias-in-coreference-resolution-on-winobias.md?metric=P(Other)>): Lower is better; unit: PERCENT.

- [Bias Gap](<https://www.sota2.com/research/sota/gender-bias-in-coreference-resolution-on-winobias.md?metric=Bias%20Gap>): Lower is better; unit: PERCENT.

## Recorded results

| Method | P(Stereo) | Result date | Report paper |
| --- | --- | --- | --- |
| Self-Debiasing | 49.49 | Not recorded | 69842418a8b01e76cc8ee74f |
| IG2 | 50.76 | Not recorded | 69842418a8b01e76cc8ee74f |
| BBA | 55.81 | Not recorded | 69842418a8b01e76cc8ee74f |
| Auto-Debias | 56.69 | Not recorded | 69842418a8b01e76cc8ee74f |
| FBA | 58.59 | Not recorded | 69842418a8b01e76cc8ee74f |
| Base | 61.11 | Not recorded | 69842418a8b01e76cc8ee74f |
| Prefix Prompting | 71.21 | Not recorded | 69842418a8b01e76cc8ee74f |

### Self-Debiasing

P(Stereo): 49.49. Result ID: 69964d4f5698665ec491d144.

[Reporting paper](<https://www.sota2.com/research/paper/69842418a8b01e76cc8ee74f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa721d1e209131cea5497f.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-v0.3"}. This record does not assert independent reproduction.

### IG2

P(Stereo): 50.76. Result ID: 69964d4f5698665ec491d145.

[Reporting paper](<https://www.sota2.com/research/paper/69842418a8b01e76cc8ee74f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69842418a8b01e76cc8ee74f.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-v0.3"}. This record does not assert independent reproduction.

### BBA

P(Stereo): 55.81. Result ID: 69964d4f5698665ec491d147.

[Reporting paper](<https://www.sota2.com/research/paper/69842418a8b01e76cc8ee74f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69842418a8b01e76cc8ee74f.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-v0.3"}. This record does not assert independent reproduction.

### Auto-Debias

P(Stereo): 56.69. Result ID: 69964d4f5698665ec491d142.

[Reporting paper](<https://www.sota2.com/research/paper/69842418a8b01e76cc8ee74f.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-v0.3"}. This record does not assert independent reproduction.

### FBA

P(Stereo): 58.59. Result ID: 69964d4f5698665ec491d146.

[Reporting paper](<https://www.sota2.com/research/paper/69842418a8b01e76cc8ee74f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69842418a8b01e76cc8ee74f.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-v0.3"}. This record does not assert independent reproduction.

### Base

P(Stereo): 61.11. Result ID: 69964d4f5698665ec491d141.

[Reporting paper](<https://www.sota2.com/research/paper/69842418a8b01e76cc8ee74f.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-v0.3"}. This record does not assert independent reproduction.

### Prefix Prompting

P(Stereo): 71.21. Result ID: 69964d4f5698665ec491d143.

[Reporting paper](<https://www.sota2.com/research/paper/69842418a8b01e76cc8ee74f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73311e209131cea5b0a7.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-v0.3"}. This record does not assert independent reproduction.

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