Maximum Independent Set on MIS
100Feasibility RateFALCON
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FALCONNumber of samples (N)=12026.02 | 100 | 1.18 | |
| FALCONNumber of samples (N)=82026.02 | 100 | 0.87 | |
| FALCONSampling strategy=Adaptive2026.02 | 100 | 0.84 | |
| SGECategory=LLM-based Optimization2026.02 | 94 | 3.83 | |
| LLMCoSolverNumber of samples (N)=82026.02 | 94 | 1.04 | |
| GRPOCategory=Neural CO Solver2026.02 | 83 | 1.34 | |
| LLMCoSolverNumber of samples (N)=12026.02 | 83 | 1.34 | |
| GPT-o1Category=Reasoning Model2026.02 | 82 | 8.03 | |
| SFT OnlyCategory=Neural CO Solver2026.02 | 80 | 1.71 | |
| GPT-o3-miniCategory=Reasoning Model2026.02 | 66 | 9.23 | |
| DeepSeek-R1Category=Reasoning Model2026.02 | 41 | 1.6 | |
| Qwen2.5-72BEvaluation protocol=Zero-shot2026.02 | 14 | 29.56 | |
| Claude-3.5-SonnetEvaluation protocol=Zero-shot2026.02 | 13 | 12.51 | |
| GPT-4oEvaluation protocol=Zero-shot2026.02 | 8 | 11.7 | |
| Llama-3.3-70BEvaluation protocol=Zero-shot2026.02 | 8 | 37.12 | |
| OPROCategory=LLM-based Optimization2026.02 | 7 | 5.95 | |
| DeepSeek-V3Evaluation protocol=Zero-shot2026.02 | 5 | 12.05 | |
| LMEACategory=LLM-based Optimization2026.02 | 5 | 25 | |
| PHPCategory=LLM-based Optimization2026.02 | 5 | 11.67 | |
| Claude-3.5-HaikuEvaluation protocol=Zero-shot2026.02 | 2 | 23.33 | |
| GPT-4o-miniEvaluation protocol=Zero-shot2026.02 | 0 | 0 |