Translation Error Analysis on 1,700-passage Full
224High-drift Instances CountGPT-5.5
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GPT-5.5Translator=GPT-5.52026.05 | 224 | 1.06 | 0.53 | 0.28 | 1.59 | |
| Gemini 3.1 ProTranslator=Gemini 3.1 Pro2026.05 | 257 | 2.12 | 1.06 | 0.67 | 3.18 | |
| Claude Sonnet 4.6Translator=Claude Sonnet 4.62026.05 | 302 | 2.47 | 1.53 | 1.05 | 4 | |
| Grok 4.3Translator=Grok 4.32026.05 | 420 | 4.29 | 6.82 | 5.72 | 11.12 |