Junliang Ye | 叶俊良

I am a third-year master's student in the Department of Computer Science at Tsinghua University , advised by Prof. Jun Zhu. In 2022, I obtained my B.S. in the School of Mathematical Sciences at Peking University.

My research interests lie in computer vision (e.g., 3D AIGC and video generation), multimodal large models (e.g., native large models), and reinforcement learning from human feedback (DPO, GRPO).

My email:yejl23@mails.tsinghua.edu.cn

Email  /  CV  /  Google Scholar  /  Github

Preprints

*equal contribution     †Project leader

We propose DeepMesh-v2, which generates meshes with intricate details and precise topology, surpassing state-of-the-art methods in both precision and quality.

We introduce Part-X-MLLM, a native 3D multimodal large language model that unifies diverse 3D tasks by formulating them as programs in a structured, executable grammar.

we propose Nano3D, a training-free framework for precise and coherent 3D object editing without masks.

Selected Research

We propose ​ShapeLLM-Omni, a multimodal large model that integrates ​3D generation, understanding, and editing capabilities.

We propose DeepMesh, which generates meshes with intricate details and precise topology, surpassing state-of-the-art methods in both precision and quality.

We present a comprehensive framework, coined DreamReward++, where we introduce a reward-aware noise sampling strategy to unleash text-driven diversity during the generation process while ensuring human preference alignment. Grounded by theoretical proof and extensive experiment comparisons, our method successfully generates high-fidelity and diverse 3D results with significant boosts in prompt alignment with human intention. Our results demonstrate the great potential for learning from human feedback to improve 3D generation.

We present a comprehensive framework, coined DreamReward, to learn and improve text-to-3D models from human preference feedback.

We propose ANIMATABLEDREAMER, a framework with the capability to generate generic categories of non-rigid 3D models.

Additional Projects

In this paper, we propose ReconX, a novel 3D scene reconstruction paradigm that reframes the ambiguous reconstruction challenge as a temporal generation task. The key insight is to unleash the strong generative prior of large pre-trained video diffusion models for sparse-view reconstruction.

We achieved first place in the TRECVID 2022 competition.

Website Template