LoftUp: Learning a Coordinate-Based Feature Upsampler for Vision Foundation Models Paper • 2504.14032 • Published Apr 18 • 7