List your company

Preference Alignment on Argilla-7k (test)

9.23LC Win RateMixDPO
2.64684.35596.0657.7741Feb 1, 2026

Evaluation results

MethodLinks
MixDPOBase Model=LLaMA-3-8B2026.029.2320.62
SimPOBase Model=LLaMA-3-8B2026.027.075.84
BASE SFTBase Model=LLaMA-3-8B2026.023.7310.19
SELECTIVEDPOBase Model=LLaMA-3-8B2026.023.595.22
VANILLA DPOBase Model=LLaMA-3-8B2026.022.912.17