“Weights & Biases Weave Evaluations enables users to validate agent performance at scale across curated datasets of historical events with known outcomes.”