ResearchBenchmarksQuestion Answering on GPQA (Selection Accuracy)Follow56.7Selection AccuracyGMM27.184834.847442.5150.1726Apr 20, 2026Evaluation ResultsMethodMethodLinksSelection AccuracyGMMSize=70B, Selection=Be...Size=70B, Selection=Best-of-1002026.0456.7DCLSize=70B, Selection=Be...Size=70B, Selection=Best-of-1002026.0456.01DSSize=70B, Selection=Be...Size=70B, Selection=Best-of-1002026.0454.27GMMSize=8B, Selection=Bes...Size=8B, Selection=Best-of-1002026.0437.09DSSize=8B, Selection=Bes...Size=8B, Selection=Best-of-1002026.0431.68DCLSize=8B, Selection=Bes...Size=8B, Selection=Best-of-1002026.0428.32