ResearchBenchmarksQuestion Answering on MC-QAFollow42.3Average AccuracyCGLS38.254439.304740.35541.4053Jun 13, 2025Evaluation ResultsMethodMethodLinksAverage AccuracyCGLSBackbone=Llama-3.2-1B,...Backbone=Llama-3.2-1B, Token Budget=20B2025.0642.3BL-CBackbone=Llama-3.2-1B,...Backbone=Llama-3.2-1B, Token Budget=20B2025.0639.72BL-RBackbone=Llama-3.2-1B,...Backbone=Llama-3.2-1B, Token Budget=20B2025.0638.41