Multi-objective preference selection on OpenAI models Setup A2: K=7, M*=10 benchmark leaderboard · SOTA2 Research

Multi-objective preference selection on OpenAI models Setup A2: K=7, M*=10

0.035Cost ($)CUPID

Evaluation Results

Method	Links
CUPID2026.05		0.035	1	100
CUPID without cost penalty2026.05		0.037	1	100
CUPID without language2026.05		0.04	1.6	100
Cost-constrained LMA2026.05		0.042	2.6	100
LMA2026.05		0.046	2.6	100
Latent Dueling Bandits2026.05		0.049	1	100
ϵ-greedy2026.05		0.053	1.8	100
RUCB2026.05		0.116	1	33.3