ResearchBenchmarksMulti-hop Question Answering on HotpotQA tool-augmented 1 (test)Follow72.8EMFLOWBOT36.60846.00455.464.796Apr 29, 2026Evaluation ResultsMethodMethodLinksEMFLOWBOTModel=GPT-4.1 miniModel=GPT-4.1 mini2026.0472.8GEPAModel=GPT-4.1 miniModel=GPT-4.1 mini2026.0469GEPA+MergeModel=GPT-4.1 miniModel=GPT-4.1 mini2026.0465.67TextGradModel=GPT-4.1 miniModel=GPT-4.1 mini2026.0462.33TraceModel=GPT-4.1 miniModel=GPT-4.1 mini2026.0460.33MIPROv2Model=GPT-4.1 miniModel=GPT-4.1 mini2026.0458BaselineModel=GPT-4.1 miniModel=GPT-4.1 mini2026.0438