Span-level machine translation error detection on MQM EN-ZH annotations 2024 (test)
35.95PrecisionMQM #2
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MQM #2type=human evaluator2026.03 | 35.95 | 25.68 | 29.96 | |
| Sonnet 4.5version=4.52026.03 | 34.31 | 21.57 | 26.49 | |
| MQM #1type=human evaluator2026.03 | 29.88 | 25.49 | 27.51 | |
| Haiku 4.5version=4.52026.03 | 29.88 | 15.15 | 20.11 | |
| gpt-oss 120bparameters=120b2026.03 | 28.89 | 21.42 | 24.6 | |
| Qwen3 235bparameters=235b2026.03 | 28.88 | 30.92 | 29.87 |