List your company
ResearchFormerly WizWand

Personalized Generation on BESPOKE

9.94R-LP-CHECK
7.31927.99968.689.3604Jan 6, 2026
Updated 8mo ago

Evaluation results

MethodLinks
P-CHECKAlign. Method=DPO, Policy π=Llama3-8b2026.019.949.6761.21
P-CHECKAlign. Method=Best-of-N Selection (BoN), Policy π=GPT-4o-mini2026.019.618.4757.32
P-CHECKAlign. Method=Best-of-N Selection (BoN), Policy π=Llama3-8b2026.019.438.2259.76
VPLAlign. Method=DPO, Policy π=Llama3-8b2026.018.687.753.55
VPLAlign. Method=Best-of-N Selection (BoN), Policy π=GPT-4o-mini2026.018.457.4953.36
CoT distillAlign. Method=DPO, Policy π=Llama3-8b2026.018.417.6356.64
PALAlign. Method=Best-of-N Selection (BoN), Policy π=Llama3-8b2026.018.327.151.68
PALAlign. Method=Best-of-N Selection (BoN), Policy π=GPT-4o-mini2026.018.317.4652.03
CoT distillAlign. Method=Best-of-N Selection (BoN), Policy π=Llama3-8b2026.018.247.1454.9
CoT distillAlign. Method=Best-of-N Selection (BoN), Policy π=GPT-4o-mini2026.018.247.6254.76
VPLAlign. Method=Best-of-N Selection (BoN), Policy π=Llama3-8b2026.018.057.0652.19
SynthMeAlign. Method=Best-of-N Selection (BoN), Policy π=GPT-4o-mini2026.018.016.4652.96
SynthMeAlign. Method=Best-of-N Selection (BoN), Policy π=Llama3-8b2026.017.996.8552.06
Default PolicyAlign. Method=Best-of-N Selection (BoN), Policy π=Llama3-8b2026.017.926.0951.55
Default PolicyAlign. Method=DPO, Policy π=Llama3-8b2026.017.926.0951.55
Default PolicyAlign. Method=Best-of-N Selection (BoN), Policy π=GPT-4o-mini2026.017.836.4351.69
SynthMeAlign. Method=DPO, Policy π=Llama3-8b2026.017.746.4552.46
PALAlign. Method=DPO, Policy π=Llama3-8b2026.017.426.1153.49