Quality Estimation on EN-HI Gender-explicit
79.1AccuracyFairQE (ours, w/ MetricX 24 L)
Evaluation Results
| Method | Links | |
|---|---|---|
| FairQE (ours, w/ MetricX 24 L)Backbone=MetricX 24 L2026.04 | 79.1 | |
| MetricX 24 XLBackbone=MetricX, Model Scale=24 XL2026.04 | 78.9 | |
| MetricX 24 LBackbone=MetricX, Model Scale=24 L2026.04 | 74 | |
| FairQE (ours, w/ COMETKiwi 22)Backbone=COMETKiwi 222026.04 | 74 | |
| COMETKiwi 23 XLBackbone=COMETKiwi, Model Scale=23 XL2026.04 | 73.1 | |
| GEMBA-MQMBackbone=GEMBA-MQM2026.04 | 72 | |
| COMETKiwi 22Backbone=COMETKiwi, Model Scale=222026.04 | 55.3 |