ResearchBenchmarksKnowledge Evaluation on PopQA (Evaluation)Follow11.2AccuracyGAME-LoRA10.99211.04611.111.154Jan 31, 2026Evaluation ResultsMethodMethodLinksAccuracyGAME-LoRAProtocol=Training-timeProtocol=Training-time2026.0111.2CADProtocol=Inference-timeProtocol=Inference-time2026.0111.2ActDecProtocol=Inference-timeProtocol=Inference-time2026.0111.2BaselineArchitecture=LoRA, Los...Architecture=LoRA, Loss=Vanilla CE, Protocol=Training-time2026.0111.1DisagreementProtocol=Training-timeProtocol=Training-time2026.0111.1MEProtocol=Training-timeProtocol=Training-time2026.0111