Multi-agent Selection (Pairwise Resolution) on HumanEval 19 models (test)
80.9Pairwise ResolutionMARGIN
Evaluation Results
| Method | Links | |
|---|---|---|
| MARGINSelection Strategy=Verbalized2026.05 | 80.9 | |
| Raw verbalized confidenceSelection Strategy=Raw verbalized2026.05 | 59.1 |