Instruction Following on IFEval (Strict/Loose Accuracy)
35.12Strict AccuracyCPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CPOBackbone=Llama-3-8B-Instruct2026.05 | 35.12 | 43.99 | |
| RDPOBackbone=Llama-3-8B-Instruct2026.05 | 34.57 | 43.62 | |
| DPOBackbone=Llama-3-8B-Instruct2026.05 | 34.01 | 40.67 | |
| SimPOBackbone=Llama-3-8B-Instruct2026.05 | 33.83 | 42.81 | |
| Llama-8B-InstructModel Size=8B, Series=Llama 32026.05 | 32.35 | 39.56 |