Supervised Fine-Tuning on SFT Mixture 99 tasks (test)
85.3Textual EntailmentDoReMi
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| DoReMiModel=500M Qwen-22026.06 | 85.3 | 74.9 | 83.6 | 78.1 | 87 | 79 | 0.249 | 81.6 | |
| TANDEMModel=500M Qwen-22026.06 | 85.3 | 76.3 | 86.2 | 78.6 | 88.5 | 84.9 | 0.208 | 83.3 | |
| UniformModel=500M Qwen-22026.06 | 84.4 | 75.1 | 86.2 | 77.8 | 88.3 | 83.2 | 0.231 | 82.5 | |
| Skill-ItModel=500M Qwen-22026.06 | 84.1 | 75 | 86.4 | 78.3 | 87.9 | 83.7 | 0.232 | 82.6 | |
| AioliModel=500M Qwen-22026.06 | 83.9 | 75.8 | 86.2 | 78 | 88.3 | 84 | 0.229 | 82.7 | |
| DoGEModel=500M Qwen-22026.06 | 81.4 | 76.7 | 86.5 | 78.6 | 88.3 | 82.4 | 0.297 | 82.4 |