ResearchBenchmarksMulti-hop Reasoning on StrategyQA (ASR/HSR)Follow88.4ASRShadowCoT72.876.8580.984.95Apr 8, 2025Evaluation ResultsMethodMethodLinksASRHSRShadowCoTTarget Model=LLaMATarget Model=LLaMA2025.0488.481.2DarkMindTarget Model=LLaMATarget Model=LLaMA2025.0479.671.5SABERTarget Model=LLaMATarget Model=LLaMA2025.047566.8BadChainTarget Model=LLaMATarget Model=LLaMA2025.0473.464.7