List your company

Competitive Programming on Hard Bench (HB-15) pass@5 v2

47Correlation ScoreReflexion
18.9226.2133.540.79Jun 29, 2026

Evaluation results

MethodLinks
ReflexionBackbone=Haiku-4-5, Evaluation Protocol=pass@52026.064786
AlgoSkill (MCTS)Backbone=Haiku-4-5, Variant=MCTS, Evaluation Protocol=pass@52026.0647100
CoTBackbone=Sonnet-4-5, Evaluation Protocol=pass@52026.064083
AlgoSkill-GBackbone=Sonnet-4-5, Variant=Greedy, Evaluation Protocol=pass@52026.064083
CoTBackbone=Haiku, Evaluation Protocol=pass@52026.0627100
AlgoSkill-GBackbone=Haiku, Variant=Greedy, Evaluation Protocol=pass@52026.0627100
DirectBackbone=Sonnet-4-5, Evaluation Protocol=pass@52026.06270
Self-RefineBackbone=Haiku-4-5, Evaluation Protocol=pass@52026.0627—
DirectBackbone=Haiku, Evaluation Protocol=pass@52026.0620100