Multi-agent Selection (Pairwise Resolution) on BigCodeBench (test)
70.9Pairwise ResolutionMARGIN
Evaluation Results
| Method | Links | |
|---|---|---|
| MARGINSelection Strategy=Verbalized2026.05 | 70.9 | |
| MARGINSelection Strategy=Consistency2026.05 | 70.7 | |
| Raw verbalized confidenceSelection Strategy=Raw verbalized2026.05 | 44.8 |