# Toxicity Classification on CivilComments Hard subset - demographic identities Source: [Toxicity Classification on CivilComments Hard subset - demographic identities]() Collection index: [llms.txt]() Data updated: 2026-06-19T08:53:50.606Z Evaluation of Toxicity Classification performance specifically on the CivilComments dataset's 'Hard subset - demographic identities', reporting specialized metrics like Hard Accuracy and Hard ECE. 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 | Toxicity Classification | | Dataset (including any stated split) | CivilComments Hard subset - demographic identities | | Metric | Hard Accuracy | | Unit | PERCENT | | Direction | Higher 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 - [Hard Accuracy](): Higher is better; unit: PERCENT. - [Hard ECE](): Lower is better; unit: PERCENT. - [Hard ECE+TS](): Lower is better; unit: PERCENT. ## Recorded results | Method | Hard Accuracy | Result date | Report paper | | --- | --- | --- | --- | | Robust Filtered | 86.9 | Not recorded | 6a34cd53206542251d7c61d4 | | Robust MoE | 86.8 | Not recorded | 6a34cd53206542251d7c61d4 | | Single Expert | 86.7 | Not recorded | 6a34cd53206542251d7c61d4 | | Vanilla MoE | 86.6 | Not recorded | 6a34cd53206542251d7c61d4 | | MoCaE | 86.6 | Not recorded | 6a34cd53206542251d7c61d4 | | FGR | 85.7 | Not recorded | 6a34cd53206542251d7c61d4 | | FGR + Robust | 85.3 | Not recorded | 6a34cd53206542251d7c61d4 | ### Robust Filtered Hard Accuracy: 86.9. Result ID: 6a35034afbc373592c328831. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Robust MoE Hard Accuracy: 86.8. Result ID: 6a35034afbc373592c328830. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Single Expert Hard Accuracy: 86.7. Result ID: 6a35034afbc373592c32882c. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Vanilla MoE Hard Accuracy: 86.6. Result ID: 6a35034afbc373592c32882d. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### MoCaE Hard Accuracy: 86.6. Result ID: 6a35034afbc373592c32882e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### FGR Hard Accuracy: 85.7. Result ID: 6a35034afbc373592c32882f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### FGR + Robust Hard Accuracy: 85.3. Result ID: 6a35034afbc373592c328832. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.