LoftUp: Learning a Coordinate-Based Feature Upsampler for Vision Foundation Models Paper โข 2504.14032 โข Published Apr 18, 2025 โข 7