ResearchBenchmarksMulti-hop Question Answering on 2WikiMQA (test)Follow48.6Exact MatchQ-DREAM22.629.3536.142.85May 31, 2025Evaluation ResultsMethodMethodLinksExact MatchF1 ScoreQ-DREAMBackbone=ChatGPTBackbone=ChatGPT2025.0548.662.1IRCoTBackbone=ChatGPTBackbone=ChatGPT2025.0541.955.2SUREBackbone=ChatGPTBackbone=ChatGPT2025.0532.838.1Q-DREAMBackbone=Llama2-7BBackbone=Llama2-7B2025.0532.439.7InstructRAGBackbone=Llama3-8BBackbone=Llama3-8B2025.0530.438.9ChatQA2Backbone=Llama3-8BBackbone=Llama3-8B2025.052935.2ChatGPTBackbone=GPT-3.5-turboBackbone=GPT-3.5-turbo2025.0523.629.2