ResearchBenchmarksDirect Question Answering on PISTOL (A_B)For AI assistantsMarkdownContent indexDirect Question Answering on PISTOL (A_B)Follow79.1ROUGE-LNPO48.73256.61664.572.384Oct 20, 2025Evaluation resultsMethodMethodLinksROUGE-LLLM Judge ScoreNPOVariant=Context-awareVariant=Context-aware2025.1079.10NPOVariant=VanillaVariant=Vanilla2025.1078.80RMUVariant=Context-awareVariant=Context-aware2025.1076.510RMUVariant=VanillaVariant=Vanilla2025.1070.55UNDIALVariant=Context-awareVariant=Context-aware2025.1050.25UNDIALVariant=VanillaVariant=Vanilla2025.1049.910