Automated Algorithm Design on ASP
5.24Top-1 Relative GapFunSearch+BehaveSim
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FunSearch+BehaveSimLLM=GPT-5-Nano, Max Evaluations=10,000, Timeout=50s2026.03 | 5.24 | 5.28 | |
| EoH+BehaveSimLLM=GPT-5-Nano, Max Evaluations=10,000, Timeout=50s2026.03 | 5.33 | 5.59 | |
| FunSearchLLM=GPT-5-Nano, Max Evaluations=10,000, Timeout=50s2026.03 | 5.84 | 6.15 | |
| EoHLLM=GPT-5-Nano, Max Evaluations=10,000, Timeout=50s2026.03 | 5.99 | 6.08 | |
| FunSearchBackbone=Llama-8B, Fine-tuned=true2025.07 | 7.43 | 8.09 | |
| EoHBackbone=Llama-8B, Fine-tuned=true2025.07 | 8.19 | 8.69 | |
| FunSearchBackbone=Llama-8B, Fine-tuned=false2025.07 | 8.46 | 9.88 | |
| EoHBackbone=Llama-1B, Fine-tuned=true2025.07 | 8.52 | 8.58 | |
| EoHBackbone=Pangu-7B, Fine-tuned=true2025.07 | 8.68 | 9.66 | |
| EoHBackbone=Pangu-7B, Fine-tuned=false2025.07 | 8.81 | 9.69 | |
| EoHBackbone=Llama-8B, Fine-tuned=false2025.07 | 8.99 | 9.83 | |
| FunSearchBackbone=Llama-1B, Fine-tuned=true2025.07 | 9.19 | 9.88 | |
| FunSearchBackbone=Pangu-7B, Fine-tuned=true2025.07 | 10.94 | 11.19 | |
| FunSearchBackbone=Pangu-7B, Fine-tuned=false2025.07 | 11 | 13.35 | |
| EoHBackbone=Llama-1B, Fine-tuned=false2025.07 | 17.02 | 19.49 | |
| FunSearchBackbone=Llama-1B, Fine-tuned=false2025.07 | 19.48 | 24.22 |