Multi-task Performance Evaluation on Aggregated Benchmarks
1.3Average RankRG-OPD (ours)
Evaluation Results
| Method | Links | |
|---|---|---|
| RG-OPD (ours)Generation Length=10242026.07 | 1.3 | |
| RG-OPD (ours)Generation Length=80962026.07 | 1.3 | |
| Qwen2.5(1.5B)SGeneration Length=10242026.07 | 2.7 | |
| Reverse KL [1]Generation Length=80962026.07 | 2.7 | |
| Qwen2.5(1.5B)SGeneration Length=80962026.07 | 2.8 | |
| Reverse KL [1]Generation Length=10242026.07 | 3 | |
| TSD-KD [10]Generation Length=10242026.07 | 3 | |
| TSD-KD [10]Generation Length=80962026.07 | 3.2 |