Multi-agent Selection (Pairwise Resolution) on HumanEval consensus 10 models (test)
82.2Pairwise ResolutionMARGIN
Evaluation Results
| Method | Links | |
|---|---|---|
| MARGINSelection Strategy=Consistency2026.05 | 82.2 | |
| MARGINSelection Strategy=Verbalized2026.05 | 76 | |
| Raw verbalized confidenceSelection Strategy=Raw verbalized2026.05 | 70.2 |