ResearchBenchmarksCommonsense Reasoning on Commonsense QACommonsense Reasoning on Commonsense QAFollow3.9Average Relative ImprovementTBDF-0.24960.82771.9052.9823Jan 29, 2026Evaluation resultsMethodMethodLinksAverage Relative ImprovementInferior/Superior Case CountsTBDFFiltering Mode=General...Filtering Mode=General, Models=Gemma 3 (270M, 1B, 4B)2026.013.9—TBDFFiltering Mode=FW-EDU,...Filtering Mode=FW-EDU, Models=Gemma 3 (270M, 1B, 4B)2026.013.41—CBFiltering Mode=FW-EDU,...Filtering Mode=FW-EDU, Models=Gemma 3 (270M, 1B, 4B)2026.01-0.09—