Error Span Detection on WMT24 (test)
84.8SPALlama-MBR-SOFTF1
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Llama-MBR-SOFTF1Backbone=Llama-3.3-70B-Inst, N (Sample size)=256, Reference Protocol=Reference-based2025.12 | 84.8 | 57.1 | 93.2 | 51.3 | |
| xCOMET-RegEvaluation Mode=Regression (score-only), Reference Protocol=Reference-based2025.12 | 84.4 | 58.1 | — | — | |
| xCOMET-QE-RegEvaluation Mode=Regression (score-only), Reference Protocol=Reference-free/QE2025.12 | 82.5 | 54.9 | — | — | |
| Llama-MAPBackbone=Llama-3.3-70B-Inst, N (Sample size)=256, Reference Protocol=Reference-based2025.12 | 82.3 | 56.8 | 91.9 | 53.1 | |
| xCOMET-ESDEvaluation Mode=Token-classification, Reference Protocol=Reference-based2025.12 | 75.7 | 55.3 | 88.9 | 30.2 | |
| xCOMET-QE-ESDEvaluation Mode=Token-classification, Reference Protocol=Reference-free/QE2025.12 | 68.8 | 54.1 | 87.9 | 28.9 |