ResearchBenchmarksQuestion Answering on HotpotQA subset of 500 questions (dev)Follow31.4EM (ReAct)FireAct14.13618.61823.127.582Oct 9, 2023Evaluation ResultsMethodMethodLinksEM (ReAct)EM (FireAct)Absolute DifferenceRelative DifferenceFireActBackbone=GPT-3.5Backbone=GPT-3.52023.1031.439.27.825FireActBackbone=CodeLlama-34BBackbone=CodeLlama-34B2023.1022.227.85.625FireActBackbone=Llama-2-13BBackbone=Llama-2-13B2023.1021.234.413.162FireActBackbone=CodeLlama-13BBackbone=CodeLlama-13B2023.1020.8298.239FireActBackbone=CodeLlama-7BBackbone=CodeLlama-7B2023.1017.427.810.460FireActBackbone=Llama-2-7BBackbone=Llama-2-7B2023.1014.826.211.477