Orienteering Problem on OP
100FeasibilityLLMCoSolver
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LLMCoSolverNumber of samples (N)=82026.02 | 100 | 1.85 | |
| FALCONNumber of samples (N)=12026.02 | 100 | 2.14 | |
| FALCONNumber of samples (N)=82026.02 | 100 | 1.42 | |
| FALCONSampling strategy=Adaptive2026.02 | 100 | 1.38 | |
| SGECategory=LLM-based Optimization2026.02 | 93 | 24.49 | |
| GRPOCategory=Neural CO Solver2026.02 | 92 | 4.25 | |
| LLMCoSolverNumber of samples (N)=12026.02 | 92 | 4.25 | |
| OPROCategory=LLM-based Optimization2026.02 | 85 | 53.96 | |
| GPT-4o-miniEvaluation protocol=Zero-shot2026.02 | 80 | 70.7 | |
| DeepSeek-R1Category=Reasoning Model2026.02 | 60 | 40.54 | |
| GPT-4oEvaluation protocol=Zero-shot2026.02 | 59 | 55.19 | |
| SFT OnlyCategory=Neural CO Solver2026.02 | 54 | 2.32 | |
| DeepSeek-V3Evaluation protocol=Zero-shot2026.02 | 50 | 46.1 | |
| Claude-3.5-SonnetEvaluation protocol=Zero-shot2026.02 | 49 | 34.62 | |
| LMEACategory=LLM-based Optimization2026.02 | 48 | 66.18 | |
| PHPCategory=LLM-based Optimization2026.02 | 43 | 36.08 | |
| Qwen2.5-72BEvaluation protocol=Zero-shot2026.02 | 32 | 49.36 | |
| GPT-o1Category=Reasoning Model2026.02 | 31 | 40.9 | |
| Llama-3.3-70BEvaluation protocol=Zero-shot2026.02 | 27 | 48.98 | |
| Claude-3.5-HaikuEvaluation protocol=Zero-shot2026.02 | 26 | 51.26 | |
| GPT-o3-miniCategory=Reasoning Model2026.02 | 8 | 43.93 |