Instruction Following on IFEval (Accuracy, Delta)
84.53AccuracyQwen2.5(14B)T
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen2.5(14B)TGeneration Length=80962026.07 | 84.53 | — | |
| Qwen2.5(14B)TGeneration Length=10242026.07 | 84.17 | — | |
| TSD-KD [10]*Generation Length=10242026.07 | 55.2 | 3.08 | |
| TSD-KD [10]Generation Length=80962026.07 | 54.8 | 2.56 | |
| TSD-KD [10]Generation Length=10242026.07 | 54.68 | 2.56 | |
| Qwen2.5(1.5B)SGeneration Length=80962026.07 | 52.24 | 0 | |
| Qwen2.5(1.5B)SGeneration Length=10242026.07 | 52.12 | 0 | |
| RG-OPD (ours)Generation Length=80962026.07 | 48.68 | -3.56 | |
| RG-OPD (ours)Generation Length=10242026.07 | 48.6 | -3.52 | |
| Reverse KL [1]Generation Length=10242026.07 | 46.28 | -5.84 | |
| Reverse KL [1]Generation Length=80962026.07 | 46.08 | -6.16 |