Cost-aware Bayesian Optimization on CAF Continuous Optimization Benchmark
2.33Ackley ErrorEIpu
Evaluation Results
| Method | Links | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| EIpuBO budget=30, trials per test function=102026.05 | 2.33 | 5.62 | 0.34 | 2.36 | 0.01 | 0.12 | 0.02 | 0 | 19.83 | 7.92 | 0.03 | 0.47 | 3.25 | — | — | |
| EIBO budget=30, trials per test function=102026.05 | 2.66 | 4.74 | 0.49 | 1.26 | 0.01 | 0.05 | 0.03 | 0 | 18.89 | 7.91 | 0.03 | 0.47 | 3.05 | — | — | |
| EI-coolBO budget=30, trials per test function=102026.05 | 2.74 | 5.78 | 0.34 | 2.29 | 0.01 | 0.07 | 0.03 | 0 | 14.95 | 8.21 | 0.03 | 0.54 | 2.92 | — | — | |
| AHD AgentMethod Paradigm=RL Model, LLM Backbone=Qwen3-4B-Instruct-2507, BO budget=30, trials per test function=102026.05 | 2.88 | 1.73 | 0.35 | 2.99 | 0.01 | 0.05 | 0.49 | 0.11 | 7 | 5.02 | 0.47 | 0.4 | 1.79 | 14.2 | 0.007 | |
| EoHMethod Paradigm=LLM-based AHD, LLM Backbone=DeepSeek-V4-Flash, BO budget=30, trials per test function=102026.05 | 3.14 | 4.02 | 0.23 | 1.63 | 0.02 | 0.16 | 0.18 | 0 | 32.02 | 7.38 | 0.18 | 0.5 | 4.12 | 100 | 0.025 | |
| MCTS-AHDMethod Paradigm=LLM-based AHD, LLM Backbone=DeepSeek-V4-Flash, BO budget=30, trials per test function=102026.05 | 3.19 | 5.29 | 0.34 | 0.83 | 0.04 | 0.07 | 0.12 | 0.01 | 26.36 | 7.59 | 0.14 | 0.61 | 3.72 | 100 | 0.071 | |
| CALMMethod Paradigm=RL Model, LLM Backbone=Qwen3-4B-Instruct-2507, BO budget=30, trials per test function=102026.05 | 3.28 | 4.22 | 1.01 | 23.05 | 0.31 | 0.57 | 1.61 | 0.04 | 50.38 | 8.44 | 0.6 | 0.86 | 7.86 | 150 | 0.057 | |
| ReEvoMethod Paradigm=LLM-based AHD, LLM Backbone=DeepSeek-V4-Flash, BO budget=30, trials per test function=102026.05 | 4.82 | 5.25 | 1.09 | 11.02 | 0.23 | 0.29 | 2.13 | 0.05 | 73.56 | 8.7 | 0.61 | 0.75 | 9.04 | 100 | 0.153 |