# Stereotype Fairness Identification on WinoBias cloze-style (test) Source: [Stereotype Fairness Identification on WinoBias cloze-style (test)]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:57:14.146Z Evaluation of model performance on the test split of the WinoBias cloze-style dataset, focusing on identifying stereotypical associations using metrics like P\_stereo and 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 | Stereotype Fairness Identification | | Dataset (including any stated split) | WinoBias cloze-style (test) | | Metric | P\_stereo | | Unit | PERCENT | | Direction | Lower is better | | Results loaded | 14 of 14 | | 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](): Lower is better; unit: PERCENT. - [P\_anti](): Higher is better; unit: PERCENT. - [P\_other](): Higher is better; unit: PERCENT. - [Gap](): Lower is better; unit: PERCENT. ## Recorded results | Method | P\_stereo | Result date | Report paper | | --- | --- | --- | --- | | IG2 | 43.18 | Not recorded | 69842418a8b01e76cc8ee74f | | BBA | 50.51 | Not recorded | 69842418a8b01e76cc8ee74f | | BBA | 51.52 | Not recorded | 69842418a8b01e76cc8ee74f | | FBA | 52.53 | Not recorded | 69842418a8b01e76cc8ee74f | | IG2 | 57.32 | Not recorded | 69842418a8b01e76cc8ee74f | | Prefix Prompting | 58.08 | Not recorded | 69842418a8b01e76cc8ee74f | | FBA | 59.34 | Not recorded | 69842418a8b01e76cc8ee74f | | Auto-Debias | 61.49 | Not recorded | 69842418a8b01e76cc8ee74f | | Base | 62.63 | Not recorded | 69842418a8b01e76cc8ee74f | | Self-Debiasing | 85.61 | Not recorded | 69842418a8b01e76cc8ee74f | | Self-Debiasing | 88.64 | Not recorded | 69842418a8b01e76cc8ee74f | | Base | 95.71 | Not recorded | 69842418a8b01e76cc8ee74f | | Prefix Prompting | 96.21 | Not recorded | 69842418a8b01e76cc8ee74f | | Auto-Debias | 96.46 | Not recorded | 69842418a8b01e76cc8ee74f | ### IG2 P\_stereo: 43.18. Result ID: 69964d4f5698665ec491d153. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Llama-3.2"}. This record does not assert independent reproduction. ### BBA P\_stereo: 50.51. Result ID: 69964d4f5698665ec491d14e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Llama-3.1"}. This record does not assert independent reproduction. ### BBA P\_stereo: 51.52. Result ID: 69964d4f5698665ec491d155. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Llama-3.2"}. This record does not assert independent reproduction. ### FBA P\_stereo: 52.53. Result ID: 69964d4f5698665ec491d14d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Llama-3.1"}. This record does not assert independent reproduction. ### IG2 P\_stereo: 57.32. Result ID: 69964d4f5698665ec491d14c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Llama-3.1"}. This record does not assert independent reproduction. ### Prefix Prompting P\_stereo: 58.08. Result ID: 69964d4f5698665ec491d14a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Llama-3.1"}. This record does not assert independent reproduction. ### FBA P\_stereo: 59.34. Result ID: 69964d4f5698665ec491d154. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Llama-3.2"}. This record does not assert independent reproduction. ### Auto-Debias P\_stereo: 61.49. Result ID: 69964d4f5698665ec491d149. [Reporting paper]() Recorded evaluation configuration: {"Model":"Llama-3.1"}. This record does not assert independent reproduction. ### Base P\_stereo: 62.63. Result ID: 69964d4f5698665ec491d148. [Reporting paper]() Recorded evaluation configuration: {"Model":"Llama-3.1"}. This record does not assert independent reproduction. ### Self-Debiasing P\_stereo: 85.61. Result ID: 69964d4f5698665ec491d14b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Llama-3.1"}. This record does not assert independent reproduction. ### Self-Debiasing P\_stereo: 88.64. Result ID: 69964d4f5698665ec491d152. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Llama-3.2"}. This record does not assert independent reproduction. ### Base P\_stereo: 95.71. Result ID: 69964d4f5698665ec491d14f. [Reporting paper]() Recorded evaluation configuration: {"Model":"Llama-3.2"}. This record does not assert independent reproduction. ### Prefix Prompting P\_stereo: 96.21. Result ID: 69964d4f5698665ec491d151. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Llama-3.2"}. This record does not assert independent reproduction. ### Auto-Debias P\_stereo: 96.46. Result ID: 69964d4f5698665ec491d150. [Reporting paper]() Recorded evaluation configuration: {"Model":"Llama-3.2"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.