Fine-grained Quality Estimation on EN-DE (test)
28.58Word-level MCCRIEQE
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| RIEQEreasoning=Explicit2026.05 | 28.58 | 32.63 | 26.43 | 27.42 | 25.5 | |
| DCSQE2026.05 | 27.11 | 30.61 | 25.89 | 21.2 | 33.26 | |
| RIEQE-NonThinkingreasoning=Implicit2026.05 | 26.82 | 31.07 | 25 | 25.68 | 24.35 | |
| xComet-XXL2026.05 | 26.53 | 31.83 | 24.23 | 17.05 | 41.81 | |
| GPT-5.52026.05 | 25.07 | 28.33 | 22.82 | 14.25 | 57.32 | |
| CometKiwi-232026.05 | 21.5 | — | 23.5 | — | — | |
| Qwen3-4B-Thinking-2507subtask process=3-step evaluation2026.05 | 17.33 | 22.26 | 19.7 | 19.05 | 20.4 | |
| Qwen3-4B-Thinking-25072026.05 | 15.81 | 21.73 | 18.73 | 16.55 | 21.57 |