Word-level Fine-grained Quality Estimation on en-mr
28.68MCCRIEQE-NonThinking
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| RIEQE-NonThinkingCategory=Our Method, Thinking=false2026.05 | 28.68 | 33.61 | 41.26 | 28.36 | |
| RIEQECategory=Our Method2026.05 | 28.47 | 33.31 | 42.23 | 27.5 | |
| DCSQECategory=Baselines2026.05 | 28.06 | 34.23 | 25.17 | 53.47 | |
| CometKiwi-23Category=Baselines2026.05 | 25.2 | — | — | — | |
| GPT-5.5Category=Baselines2026.05 | 18.55 | 27.47 | 19.54 | 46.27 | |
| xComet-XXLCategory=Baselines2026.05 | 14.53 | 25.12 | 19.79 | 34.38 | |
| Qwen3-4B-Thinking-2507Category=Backbone Model2026.05 | 9.15 | 18.71 | 20.53 | 17.19 | |
| Qwen3-4B-Thinking-2507Category=Backbone Model, Subtask=true2026.05 | 4.67 | 12.02 | 14.54 | 10.24 |