Evaluations

Model performance across 818 scenarios, 9 clusters, and 3 domains. Compare frontier and fine-tuned models side by side.