kapynResearch

Fortress: A Case Study in Stabilizing Search Recommendations via Temporal Data Augmentation and Feature Pruning

Fortress is a new framework stabilizing search recommendation models by pruning volatile features. It addresses temporal instability that degrades reliability in multi-stage systems by using historical data to identify and remove inconsistent predictors, improving accuracy and user experience.

Apple ML Research·Jul 6, 2026

Opening Kapyn…