ResearchBenchmarksReasoning on CommonsenseQAFollow65.8AccuracyCES56.12858.63961.1563.661May 19, 2026Evaluation ResultsMethodMethodLinksAccuracyResponse LengthCESBackbone=DeepSeek-R1-D...Backbone=DeepSeek-R1-Distill-1.5B2026.0565.8448DAPOBackbone=DeepSeek-R1-D...Backbone=DeepSeek-R1-Distill-1.5B2026.0556.5451