ResearchBenchmarksGraduate-level Reasoning on SuperGPQAGraduate-level Reasoning on SuperGPQAFollow33.6AccuracyRealSafe31.41631.98332.5533.117Sep 29, 2025Evaluation resultsMethodMethodLinksAccuracyRealSafebackbone=DS-8B, alignm...backbone=DS-8B, alignment=SFT-based2025.0933.6Basebackbone=DS-8Bbackbone=DS-8B2025.0933.5IPObackbone=DS-8Bbackbone=DS-8B2025.0931.5