Span-level Machine Translation Error Detection on WMT MQM (EN-DE) 2022 (test)
42.66PrecisionMQM #1
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MQM #1type=human evaluator2026.03 | 42.66 | 44.62 | 43.62 | |
| MQM #2type=human evaluator2026.03 | 39.22 | 43.56 | 41.28 | |
| Sonnet 4.5version=4.52026.03 | 32.61 | 29.75 | 31.12 | |
| Haiku 4.5version=4.52026.03 | 30.07 | 19.89 | 23.94 | |
| gpt-oss 120bparameters=120b2026.03 | 24.08 | 31.04 | 27.12 | |
| Qwen3 235bparameters=235b2026.03 | 23.62 | 38.87 | 29.39 |