List your company

Multi-step reasoning and knowledge retrieval on GPQA (test)

52.27AccuracyDynamic Persona Routing
49.656550.9632552.2753.57675Apr 13, 2026

Evaluation results

MethodLinks
Dynamic Persona RoutingTotal=448, Sampled=442026.0452.27—23
Best Static BaselineTotal=448, Sampled=442026.04—41.96—