ResearchBenchmarksMulti-question Reasoning on HotpotQA 5Q (Exact Match)Multi-question Reasoning on HotpotQA 5Q (Exact Match)Follow15.4Exact Match AccuracyM2R5.1047.77710.4513.123Apr 10, 2026Evaluation resultsMethodMethodLinksExact Match AccuracyM2R2026.0415.4ReSearch2026.0413.9SURE2026.0413.1COFT2026.049.5Naive RAG2026.048.2Naive Generation2026.045.5