List your company

Personalization on Multi-Bench (MB)

94.84Win RateMIPO
12.305633.732855.1676.5872Mar 10, 2026

Evaluation results

MethodLinks
MIPOBase Model=Llama-3.2-3B-Instruct2026.0394.84
RLAIF (MI)Base Model=Llama-3.2-3B-Instruct2026.0393.65
INFONCEBase Model=Llama-3.2-3B-Instruct2026.0393.65
MIPOBase Model=Llama-3.2-1B-Instruct2026.0393.26
INFONCEBase Model=Llama-3.2-1B-Instruct2026.0391.67
SFTBase Model=Llama-3.2-3B-Instruct2026.0390.48
RLAIFBase Model=Llama-3.2-3B-Instruct2026.0390.48
PPO (MI)Base Model=Llama-3.2-3B-Instruct2026.0390.08
SFT (Revision)Base Model=Llama-3.2-3B-Instruct2026.0388.89
RLAIFBase Model=Llama-3.2-1B-Instruct2026.0384.12
Personalized PromptingBase Model=Llama-3.2-3B-Instruct2026.0383.33
RLAIF (MI)Base Model=Llama-3.2-1B-Instruct2026.0382.93
PPO (MI)Base Model=Llama-3.2-1B-Instruct2026.0382.22
Personalized PromptingBase Model=Llama-3.2-1B-Instruct2026.0379.76
SFTBase Model=Llama-3.2-1B-Instruct2026.0378.97
RLAIF (MI)Base Model=Qwen2.5-7B-Instruct2026.0378.57
SFT (Revision)Base Model=Llama-3.2-1B-Instruct2026.0377.78
Llama-3.2-3B-InstructBase Model=Llama-3.2-3B-Instruct2026.0377.38
RLAIFBase Model=Qwen2.5-7B-Instruct2026.0376.19
SFTBase Model=Qwen2.5-7B-Instruct2026.0375
MIPOBase Model=Qwen2.5-1.5B-Instruct2026.0374.6
MIPOBase Model=Qwen2.5-3B-Instruct2026.0374.6
MIPOBase Model=Qwen2.5-7B-Instruct2026.0373.81
INFONCEBase Model=Qwen2.5-3B-Instruct2026.0373.02
SFT (Revision)Base Model=Qwen2.5-7B-Instruct2026.0372.62
PPO (MI)Base Model=Qwen2.5-7B-Instruct2026.0372.62
INFONCEBase Model=Qwen2.5-7B-Instruct2026.0372.62
RLAIF (MI)Base Model=Qwen2.5-3B-Instruct2026.0372.22
RLAIFBase Model=Qwen2.5-3B-Instruct2026.0370.64
Personalized PromptingBase Model=Qwen2.5-7B-Instruct2026.0370.24
PPO (MI)Base Model=Qwen2.5-3B-Instruct2026.0366.67
SFT (Revision)Base Model=Qwen2.5-3B-Instruct2026.0364.68
SFTBase Model=Qwen2.5-3B-Instruct2026.0364.29
Personalized PromptingBase Model=Qwen2.5-3B-Instruct2026.0363.1
Llama-3.2-1B-InstructBase Model=Llama-3.2-1B-Instruct2026.0359.52
INFONCEBase Model=Qwen2.5-1.5B-Instruct2026.0359.52
Qwen2.5-7B-InstructBase Model=Qwen2.5-7B-Instruct2026.0347.62
Qwen2.5-3B-InstructBase Model=Qwen2.5-3B-Instruct2026.0346.43
SFTBase Model=Qwen2.5-1.5B-Instruct2026.0339.68
Personalized PromptingBase Model=Qwen2.5-1.5B-Instruct2026.0339.29
RLAIFBase Model=Qwen2.5-1.5B-Instruct2026.0339.29
PPO (MI)Base Model=Qwen2.5-1.5B-Instruct2026.0336.9
SFT (Revision)Base Model=Qwen2.5-1.5B-Instruct2026.0335.31
RLAIF (MI)Base Model=Qwen2.5-1.5B-Instruct2026.0334.52
Qwen2.5-1.5B-InstructBase Model=Qwen2.5-1.5B-Instruct2026.0315.48