Code Generation on MBPP (nR2, HV_opt, HV_pes, Gap)
0.429nR2Initial
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| InitialModel=GPT-OSS-120B, Token Budget=7.5M2026.05 | 0.429 | 0.278 | 0.259 | 0.019 | |
| InitialModel=Qwen3-30B, Token Budget=7.5M2026.05 | 0.38 | 0.392 | 0.331 | 0.061 | |
| NSGA-II-POModel=GPT-OSS-120B, Token Budget=7.5M2026.05 | 0.306 | 0.504 | 0.469 | 0.035 | |
| InitialModel=Mistral-3.2-24B, Token Budget=7.5M2026.05 | 0.196 | 0.731 | 0.671 | 0.06 | |
| MO-CAPOModel=Qwen3-30B, Token Budget=7.5M2026.05 | 0.165 | 0.832 | 0.73 | 0.102 | |
| MO-CAPOModel=GPT-OSS-120B, Token Budget=7.5M2026.05 | 0.144 | 0.84 | 0.834 | 0.005 | |
| NSGA-II-POModel=Qwen3-30B, Token Budget=7.5M2026.05 | 0.144 | 0.86 | 0.848 | 0.012 | |
| MO-CAPOModel=Mistral-3.2-24B, Token Budget=7.5M2026.05 | 0.107 | 0.979 | 0.859 | 0.12 | |
| NSGA-II-POModel=Mistral-3.2-24B, Token Budget=7.5M2026.05 | 0.092 | 0.995 | 0.936 | 0.059 |