Instruction Following on NoveltyBench
40.1Lexical DominanceAligned
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Aligned2025.11 | 40.1 | 27.3 | 12.8 | 17.2 | 20 | 28.6 | |
| BACO bestselection_criteria=best router2025.11 | 31 | 49.5 | 45.2 | 48.8 | 47.4 | 39.9 | |
| Base2025.11 | 9.8 | 14.2 | 14.2 | 13.1 | 14.2 | 11.4 | |
| Prompting bestselection_criteria=best in category2025.11 | 8 | — | — | 6.5 | — | 7.3 | |
| Nudging2025.11 | 6.8 | 19.2 | 16.1 | 7.6 | 17.6 | 7.2 | |
| Ensemble bestselection_criteria=best in category2025.11 | 3.4 | — | — | 5.8 | — | 4.6 | |
| Decoding2025.11 | 0.8 | — | — | 1 | — | 0.9 |