Code Generation on LCB cross-lingual
58.28RPRDPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DPOMode=Fine-tuned, English occurrence=language confusion2026.04 | 58.28 | 81.59 | |
| TLPON=16, Mode=Fine-tuned, English occurrence=language confusion2026.04 | 57.25 | 78.66 | |
| ORPOMode=Fine-tuned, English occurrence=language confusion2026.04 | 50.37 | 82.34 | |
| BaselineDescription=Original model prior to fine-tuning2026.04 | 43.88 | 78.05 | |
| SFTMode=Fine-tuned, English occurrence=language confusion2026.04 | 38.33 | 64.37 |