ResearchBenchmarksLocal Question Answering on RealMMBench TechSlidesFor AI assistantsMarkdownContent indexLocal Question Answering on RealMMBench TechSlidesFollow60.86AccuracyMegaRAG14.236826.340938.44550.5491Nov 26, 2025Evaluation resultsMethodMethodLinksAccuracyMegaRAGResponse LLM=GPT-4o-miniResponse LLM=GPT-4o-mini2025.1160.86NaiveRAGResponse LLM=GPT-4o-miniResponse LLM=GPT-4o-mini2025.1132.94LightRAGResponse LLM=GPT-4o-miniResponse LLM=GPT-4o-mini2025.1131.39GraphRAGSearch Mode=LocalSearch Mode=Local2025.1126.81GraphRAGSearch Mode=GlobalSearch Mode=Global2025.1116.03