Mathematical Reasoning on GSM8K (nR2, HV_opt, HV_pes, Gap)
38.7nR2Initial
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| InitialModel=Mistral-3.2-24B, Token Budget=7.5M2026.05 | 38.7 | 0.352 | 0.352 | 0 | |
| InitialModel=Qwen3-30B, Token Budget=7.5M2026.05 | 37.4 | 0.361 | 0.361 | 0 | |
| NSGA-II-POModel=Qwen3-30B, Token Budget=7.5M2026.05 | 34.3 | 0.399 | 0.399 | 0 | |
| MO-CAPOModel=Qwen3-30B, Token Budget=7.5M2026.05 | 34 | 0.41 | 0.406 | 0.4 | |
| InitialModel=GPT-OSS-120B, Token Budget=7.5M2026.05 | 31.9 | 0.438 | 0.426 | 1.2 | |
| NSGA-II-POModel=Mistral-3.2-24B, Token Budget=7.5M2026.05 | 12.2 | 0.92 | 0.896 | 2.4 | |
| MO-CAPOModel=Mistral-3.2-24B, Token Budget=7.5M2026.05 | 11 | 0.943 | 0.93 | 1.3 | |
| NSGA-II-POModel=GPT-OSS-120B, Token Budget=7.5M2026.05 | 9.8 | 0.973 | 0.92 | 5.3 | |
| MO-CAPOModel=GPT-OSS-120B, Token Budget=7.5M2026.05 | 3.8 | 1.15 | 1.145 | 0.5 |