Multi-objective preference selection on OpenAI models Setup A3: K=9, M*=2 benchmark leaderboard · SOTA2 Research

Multi-objective preference selection on OpenAI models Setup A3: K=9, M*=2

0.022Cost ($)CUPID without language

Evaluation Results

Method	Links
CUPID without language2026.05		0.022	2	100
ϵ-greedy2026.05		0.058	2.5	80
CUPID without cost penalty2026.05		0.066	2	100
CUPID2026.05		0.07	2	100
Cost-constrained LMA2026.05		0.097	1	20
RUCB2026.05		0.117	1	33.3
Latent Dueling Bandits2026.05		0.125	—	0
LMA2026.05		0.125	1	20