Instruction Following on In-domain
14Win RatedaDPO
Evaluation Results
| Method | Links | |
|---|---|---|
| daDPOBackbone=Qwen2.5-1.5B2025.06 | 14 | |
| dDPOBackbone=Qwen2.5-1.5B2025.06 | 0.7 | |
| Qwen2.5-7BBackbone=Qwen2.5-7B2025.06 | 0 | |
| rDPOBackbone=Qwen2.5-1.5B2025.06 | -5 | |
| dSFTBackbone=Qwen2.5-1.5B2025.06 | -7.3 | |
| rDPOBackbone=Qwen2.5-0.5B2025.06 | -48.7 | |
| daDPOBackbone=Qwen2.5-0.5B2025.06 | -55.3 | |
| Qwen2.5-1.5BBackbone=Qwen2.5-1.5B2025.06 | -61.7 | |
| dDPOBackbone=Qwen2.5-0.5B2025.06 | -63.3 | |
| dSFTBackbone=Qwen2.5-0.5B2025.06 | -66.3 | |
| Qwen2.5-0.5BBackbone=Qwen2.5-0.5B2025.06 | -75 |