ResearchBenchmarksDatabase Question Answering on RealEHR (test)Follow55.2AccuracyGATE-0.4414.00528.4542.895Jun 4, 2026Evaluation ResultsMethodMethodLinksAccuracyGATEBackbone=GPT-5.4-miniBackbone=GPT-5.4-mini2026.0655.2ReFoRCEBackbone=GPT-5.4-miniBackbone=GPT-5.4-mini2026.0648.6ReDelBackbone=GPT-5.4-miniBackbone=GPT-5.4-mini2026.0644.8ReActBackbone=GPT-5.4-miniBackbone=GPT-5.4-mini2026.0642.9GATEBackbone=Qwen3-8BBackbone=Qwen3-8B2026.065.6ReFoRCEBackbone=Qwen3-8BBackbone=Qwen3-8B2026.063.8ReDelBackbone=Qwen3-8BBackbone=Qwen3-8B2026.062.5ReActBackbone=Qwen3-8BBackbone=Qwen3-8B2026.061.7