WWDC26: Improve your prompts by hill-climbing with Evaluations | Apple
Apple's new Evaluations framework lets developers hill-climb and align model judges to reduce drift in AI features.
“This discrepancy between model and human is known as drift, and it is a problem faced by all developers trying to evaluate intelligent features.”
Apple introduced an Evaluations framework at WWDC26 that helps developers systematically improve intelligence-powered app features through a hill-climbing loop and Xcode evaluation reports. It addresses 'drift'—the divergence between model-judge and human expert ratings—by letting developers align judges to expert opinion, a core challenge in shipping reliable AI features.