ResearchBenchmarksExact Match on HotPotQA (Multi-hop Question Answering)Follow36Exact Match (HotPotQA)ReAct0.649.821928.18Apr 5, 2026Evaluation ResultsMethodMethodLinksExact Match (HotPotQA)ReActModel=GPT-5.4Model=GPT-5.42026.0436ReActModel=Claude Haiku 4.5Model=Claude Haiku 4.52026.0426.5ReActModel=Claude Sonnet 4.6Model=Claude Sonnet 4.62026.0416.7PTRModel=GPT-5.4Model=GPT-5.42026.0412.2PTRModel=GPT-4o-MiniModel=GPT-4o-Mini2026.0412ReActModel=GPT-4o-MiniModel=GPT-4o-Mini2026.048PTRModel=Claude Sonnet 4.6Model=Claude Sonnet 4.62026.048PTRModel=Claude Haiku 4.5Model=Claude Haiku 4.52026.042