ResearchBenchmarksLabeling Task Performance on 80-session Set (4-fold CV)Follow85Fold 0 PerformanceCodex70.4474.227881.78Mar 28, 2026Evaluation ResultsMethodMethodLinksFold 0 PerformanceFold 1 PerformanceFold 2 PerformanceFold 3 PerformanceMean PerformancePerformance Standard DeviationCodexIteration strategy=Age...Iteration strategy=Agent-driven iteration2026.03856188797812ClaudeIteration strategy=Age...Iteration strategy=Agent-driven iteration2026.0371818466768GeminiIteration strategy=Age...Iteration strategy=Agent-driven iteration2026.03716788637211