Multi-agent Selection (Pairwise Resolution) on CodeContests (test)
89.4Pairwise ResolutionMARGIN
Evaluation Results
| Method | Links | |
|---|---|---|
| MARGINSelection Strategy=Consistency2026.05 | 89.4 | |
| MARGINSelection Strategy=Verbalized2026.05 | 81 | |
| Raw verbalized confidenceSelection Strategy=Raw verbalized2026.05 | 55.5 |