Optimization Model Repair on OptiRepair-SC 284 problems (test)
97.2P1 RRQwen3-8B (OptiSTaR)
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Qwen3-8B (OptiSTaR)Arch.=Qwen, Steps=5.2, Tokens=3,4392026.02 | 97.2 | 81.7 | 64.5 | 61.3 | 100 | |
| Llama-3.1-8B (OptiSTaR)Arch.=Llama, Steps=4.3, Tokens=3,1802026.02 | 97.2 | 80.3 | 67.7 | 74.2 | 100 | |
| Best API (GPT-5.2)Arch.=N/A, Steps=14.0, Tokens=20,7302026.02 | 53.5 | 42.2 | 41.9 | 38.7 | 100 | |
| Avg. API (22 models)Arch.=N/A, Steps=7.8, Tokens=30,0782026.02 | 27.6 | 21.3 | 3.7 | 4.7 | 87.3 |