ResearchBenchmarksunsound inference on Hiring Assistant val batch seed 42 (test)Follow1Mean KappaTextGrad0.944880.959190.97350.98781May 25, 2026Evaluation ResultsMethodMethodLinksMean KappaStd Dev KappaMean Gap (Dev vs Test)TextGradOptimizer=GPT-5.4Optimizer=GPT-5.42026.05100.1SPEAROptimizer=GPT-5.4Optimizer=GPT-5.42026.050.9790.020.08GEPAOptimizer=GPT-5.4Optimizer=GPT-5.42026.050.9470.020.14