Scientific Reasoning on GPQA-D (Acc, Time, Token)
66.2AccuracyRecursiveMAS
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| RecursiveMASRecursion Round=r=3, Setting=Scaled2026.04 | 66.2 | 2,638 | 2,524 | |
| RecursiveMASRecursion Round=r=2, Setting=Scaled2026.04 | 64.6 | 2,342 | 2,521 | |
| RecursiveMASRecursion Round=r=1, Setting=Scaled2026.04 | 63.1 | 1,965 | 2,675 | |
| Recursive-TextMASRecursion Round=r=1, Setting=Scaled2026.04 | 61.5 | 2,190 | 3,693 | |
| Recursive-TextMASRecursion Round=r=2, Setting=Scaled2026.04 | 59.1 | 4,207 | 6,128 | |
| Recursive-TextMASRecursion Round=r=3, Setting=Scaled2026.04 | 58.6 | 7,537 | 8,091 | |
| RecursiveMASRecursion Round=r=3, Setting=Light2026.04 | 32.6 | 861 | 786 | |
| RecursiveMASRecursion Round=r=2, Setting=Light2026.04 | 32.3 | 752 | 813 | |
| RecursiveMASRecursion Round=r=1, Setting=Light2026.04 | 30.3 | 586 | 829 | |
| Recursive-TextMASRecursion Round=r=2, Setting=Light2026.04 | 28.7 | 1,825 | 3,708 | |
| Recursive-TextMASRecursion Round=r=3, Setting=Light2026.04 | 28.7 | 3,322 | 5,820 | |
| Recursive-TextMASRecursion Round=r=1, Setting=Light2026.04 | 28.1 | 1,056 | 2,084 |