Multi-Fidelity Bayesian Optimization on MFBO environment
8.035Distance to OptimumOracle Only
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Oracle OnlyOracle Oϕ=gpt-5, Mode=Oracle Only2026.05 | 8.035 | 144 | 100 | 8.25 | |
| Oracle OnlyOracle Oϕ=gpt-5-nano, Mode=Oracle Only2026.05 | 10.41 | 149 | 100 | 2.1 | |
| Hierarchical Prompt-Domain ControlOracle Oϕ=gpt-5, Adaptive Uθ=Llama-3.1-8B, Mode=Hierarchical2026.05 | 12.74 | 121 | 3.471 | 2.25 | |
| Hierarchical Prompt-Domain ControlOracle Oϕ=gpt-5-nano, Adaptive Uθ=Llama-3.1-8B, Mode=Hierarchical2026.05 | 13.05 | 132 | 4.86 | 0.45 | |
| Hierarchical Prompt-Domain ControlOracle Oϕ=gpt-5, Adaptive Uθ=Mistral-7B, Mode=Hierarchical2026.05 | 71.92 | 23 | 11.11 | 6.12 | |
| Hierarchical Prompt-Domain ControlOracle Oϕ=gpt-5-nano, Adaptive Uθ=Mistral-7B, Mode=Hierarchical2026.05 | 84.86 | 70 | 6.25 | 0.75 | |
| Distill onlyOracle Oϕ=gpt-5, Adaptive Uθ=Llama-3.1-8B, Mode=Distill only2026.05 | 124.1 | 74 | — | 0 | |
| Distill onlyOracle Oϕ=gpt-5, Adaptive Uθ=Mistral-7B, Mode=Distill only2026.05 | 245.6 | 9 | — | 0 | |
| No distillOracle Oϕ=gpt-5, Adaptive Uθ=Llama-3.1-8B, Mode=No distill2026.05 | 435.1 | 12 | — | 0 | |
| No distillOracle Oϕ=gpt-5, Adaptive Uθ=Mistral-7B, Mode=No distill2026.05 | 464.8 | 3 | — | 0 |