Lens: Rethinking Training Efficiency for Foundational Text-to-Image Models Paper • 2605.21573 • Published 15 days ago • 107