Preference Alignment on Argilla-7k (test)
9.23LC Win RateMixDPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MixDPOBase Model=LLaMA-3-8B2026.02 | 9.23 | 20.62 | |
| SimPOBase Model=LLaMA-3-8B2026.02 | 7.07 | 5.84 | |
| BASE SFTBase Model=LLaMA-3-8B2026.02 | 3.73 | 10.19 | |
| SELECTIVEDPOBase Model=LLaMA-3-8B2026.02 | 3.59 | 5.22 | |
| VANILLA DPOBase Model=LLaMA-3-8B2026.02 | 2.9 | 12.17 |