Language Adherence on MMMLU (target)
99.49RPRTLPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| TLPOEvaluation Setting=English as a Neutral Category, N=162026.04 | 99.49 | 99.96 | |
| SFTEvaluation Setting=English as a Neutral Category2026.04 | 99.11 | 99.75 | |
| DPOEvaluation Setting=English as a Neutral Category2026.04 | 98.94 | 99.43 | |
| BaselineEvaluation Setting=English as a Neutral Category2026.04 | 98.15 | 99.89 | |
| ORPOEvaluation Setting=English as a Neutral Category2026.04 | 97.85 | 99.82 |