Fine-grained Quality Estimation zh-en (test)
30.73Word-level MCCRIEQE
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| RIEQEreasoning=Explicit2026.05 | 30.73 | 35.2 | 29.17 | 24.07 | 37 | |
| RIEQE-NonThinkingreasoning=Implicit2026.05 | 29.03 | 34.17 | 29.34 | 24.19 | 37.27 | |
| DCSQE2026.05 | 28.12 | 28.61 | 27.71 | 22.19 | 36.88 | |
| CometKiwi-232026.05 | 26.9 | — | 27.2 | — | — | |
| GPT-5.52026.05 | 25.2 | 26.43 | 20.62 | 12.08 | 70.5 | |
| Qwen3-4B-Thinking-2507subtask process=3-step evaluation2026.05 | 23.43 | 29.1 | 23.84 | 17.46 | 37.55 | |
| xComet-XXL2026.05 | 22.91 | 27.93 | 21.96 | 14.27 | 47.59 | |
| Qwen3-4B-Thinking-25072026.05 | 21.95 | 27.56 | 22.4 | 15.51 | 40.31 |