PhD student, Stanford University
1 paper at NeurIPS 2025
Rectified Point Flow is a single generative model that turns unaligned point clouds into assembled shapes—unifying pairwise registration and multi-part assembly—and sets new state-of-the-art results across five benchmarks.