Full Professor, Tsinghua University
3 papers at NeurIPS 2025
We trained MeshLLM, an Object-to-Code inference framework that generates Blender Python scripts to reconstruct 3D meshes from point clouds in a structured and editable manner.
This paper introduces the concept of co-adaptation in 3D Gaussian Splatting, analyzes its impact on rendering artifacts, and proposes strategies to reduce it.
To train an adaptive video tokenizer, we introduce probabilistic taildrop to inject visual complexity prior to the tokenizer and incorporate GRPO for post-training, which further boosts efficiency in a task-aware adaptive manner.
