PhD student, University of Science and Technology of China
1 paper at NeurIPS 2025
We introduce Mesh-RFT, a novel framework leveraging localized preference optimization with topology-aware scoring to achieve state-of-the-art high-fidelity 3D mesh generation.
