ResearchBenchmarksCompetitive Programming on CP-275 audited version 2Follow80.7Correctness (%)AlgoSkill8.83627.49346.1564.807Jun 29, 2026Evaluation ResultsMethodMethodLinksCorrectness (%)AlgoSkillBackbone=Claude 3 HaikuBackbone=Claude 3 Haiku2026.0680.7CoTBackbone=Claude 3 HaikuBackbone=Claude 3 Haiku2026.0679.3DirectBackbone=Claude 3 HaikuBackbone=Claude 3 Haiku2026.0678.2AlgoSkillBackbone=Gemini-2.5-FlashBackbone=Gemini-2.5-Flash2026.0618.2CoTBackbone=Gemini-2.5-FlashBackbone=Gemini-2.5-Flash2026.0616DirectBackbone=Gemini-2.5-FlashBackbone=Gemini-2.5-Flash2026.0611.6