Instruction Following on IFEval (Acc_prompt)
19.6Accuracy (Prompt)FiMi-RM
Evaluation Results
| Method | Links | |
|---|---|---|
| FiMi-RMBackbone=Gemma2-9B, Optimization Algorithm=DPO2025.05 | 19.6 | |
| FiMi-RMBackbone=Qwen2.5-7B, Optimization Algorithm=DPO2025.05 | 19 | |
| ODINBackbone=Qwen2.5-7B, Optimization Algorithm=DPO2025.05 | 18.6 | |
| Length PenaltyBackbone=Gemma2-9B, Optimization Algorithm=DPO2025.05 | 18.2 | |
| Vanilla RMBackbone=Gemma2-9B, Optimization Algorithm=DPO2025.05 | 18 | |
| ODINBackbone=Gemma2-9B, Optimization Algorithm=DPO2025.05 | 17.8 | |
| Vanilla RMBackbone=Qwen2.5-7B, Optimization Algorithm=DPO2025.05 | 17.4 | |
| Length PenaltyBackbone=Qwen2.5-7B, Optimization Algorithm=DPO2025.05 | 16.6 |