ResearchBenchmarksHierarchical Summarization on SummaryBench Romeo and Juliet (R&J)Hierarchical Summarization on SummaryBench Romeo and Juliet (R&J)Follow100Fidelity Rate (FR)Magentic-One-4235077May 14, 2026Evaluation resultsMethodMethodLinksFidelity Rate (FR)Word Count (WC)Magentic-One2026.05100—MegaAgent2026.0560472APWAvariant=mini x nanovariant=mini x nano2026.051185Direct2026.05019APWAvariant=mini x minivariant=mini x mini2026.05096APWAvariant=5.4 x minivariant=5.4 x mini2026.050168APWAvariant=5.4 x nanovariant=5.4 x nano2026.050174