List your company

Multi-agent Selection (Pairwise Resolution) on BigCodeBench (test)

70.9Pairwise ResolutionMARGIN
43.75650.80357.8564.897May 21, 2026

Evaluation results

MethodLinks
MARGINSelection Strategy=Verbalized2026.0570.9
MARGINSelection Strategy=Consistency2026.0570.7
Raw verbalized confidenceSelection Strategy=Raw verbalized2026.0544.8