Yuan Li

PhD Student, UT Dallas · Geometry Processing · 3D Generation

I am a PhD student at the University of Texas at Dallas, advised by Prof. Xiaohu Guo.

My work pursues one goal: building a static 3D world inside computers at the lowest possible cost. That leads me to adaptive representations — tokenizations, partitions and primitive decompositions that spend detail only where geometry actually needs it.

My recent work spans adaptive 3D tokenization for generation, CAD reconstruction from single images, and long-horizon agents that drive real DCC toolchains.

Before Dallas I received my MSc from the CAD&CG State Key Lab at Zhejiang University, and my BSc from Central South University.

News

Sep 2026SuperVoxelGPT is accepted to ECCV 2026 🎉 — adaptive and ordered 3D tokenization for autoregressive shape generation. (first author)
May 2026Started a research scientist internship at LightSpeed Studios, building a long-horizon 3D generation agent.
Dec 2025FUSE is accepted to AAAI 2026, in collaboration with Tongyi Lab.
Oct 2025Efficient ToothAR is accepted to IEEE BIBM 2025 🎉
Mar 2025CADDreamer is accepted to CVPR 2025 as a Highlight 🎉 — CAD object generation from single-view images. (first author)
Dec 2024SIGraph is accepted to AAAI 2025 🎉 (co-first author)
Dec 2023Started my PhD at UT Dallas with Prof. Xiaohu Guo.
Mar 2022MeshFormer published in Computer Graphics Forum 🎉 (first author)

Publications

ECCV 2026

First Author

SuperVoxelGPT: Adaptive and Ordered 3D Tokenization for Autoregressive Shape Generation

Yuan Li, Congyi Zhang, Xifeng Gao, Xiaohu Guo

European Conference on Computer Vision (ECCV), 2026 · with LightSpeed Studios

Partitions a shape into adaptive, ordered SuperVoxels to make autoregressive generation tractable at high resolution, cutting the token sequence by over 80%.

Paper Project Code Video

CVPR 2025 ★ Highlight

First Author

CADDreamer: CAD Object Generation from Single-view Images

Yuan Li, Cheng Lin, Yuan Liu, Xiaoxiao Long, Chenxu Zhang, Ningna Wang, Xin Li, Wenping Wang, Xiaohu Guo

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025 · Highlight

Fits primitives to single-view predictions under a watertight constraint, so the result is an editable CAD model rather than a raw mesh.

Paper Project Code Video

AAAI 2025

Co-First Author

SIGraph: Saliency Image-Graph Network for Retinal Disease Classification in Fundus Image

Peng Zhang*, Yuan Li*, Haotian Song, Yankai Jiang, Yubo Tao, Hai Lin, Hongguang Cui

AAAI Conference on Artificial Intelligence, 2025 · * equal contribution

A new graph pooling scheme keeps fine-grained lesion detail through downsampling, improving retinal disease classification on fundus images.

AAAI 2026

FUSE: Fine-Grained and Semantic-Aware Learning for Unified Image Understanding and Generation

Peng Zhang, Wanggui He, Mushui Liu, Wenyi Xiao, Siyu Zou, Yuan Li, et al.

AAAI Conference on Artificial Intelligence, 2026 · with Tongyi Lab

A coarse-to-fine GRPO post-training strategy aligns understanding with generation in a single model, improving both tasks.

BIBM 2025

Efficient ToothAR: Cascade Autoregressive Orthodontic Treatment Planning

Zhenhao Peng, Peng Zhang, Yuan Li, Hanxiao Huang, Bin Zhang, Haotian Song, Meng Geng, Yubo Tao, Hai Lin

IEEE International Conference on Bioinformatics and Biomedicine (BIBM), 2025

Casts orthodontic planning as cascaded autoregression, predicting the whole treatment sequence in one pass instead of costly step-by-step optimization.

Under Review

First Author

MSFormer: A Skeleton-multiview Fusion Method For Tooth Instance Segmentation

Yuan Li, Huan Liu, Yubo Tao, Xiangyang He, Haifeng Li, Xiaohu Guo, Hai Lin

Major revision, IEEE Transactions on Medical Imaging

Fuses skeleton priors with multi-view features to disambiguate touching teeth, giving accurate instance-level tooth segmentation.

CGF 2022

First Author

MeshFormer: High-resolution Mesh Segmentation with Graph Transformer

Yuan Li, Xiangyang He, Yankai Jiang, Huan Liu, Yubo Tao, Hai Lin

Computer Graphics Forum (Pacific Graphics), 2022

Runs a graph transformer directly on dense meshes to capture long-range context, segmenting million-face models at full resolution.

Paper Code

Experience

May – Aug 2026 Research Scientist Intern — LightSpeed Studios
Long-horizon 3D generation agent. Going from a text prompt to a production-ready game asset needs hundreds of chained calls to generative models and geometry algorithms — beyond conventional pipelines. I built the knowledge base and skill library the agent plans and acts over. Supervised by Prof. Xifeng Gao.
May – Aug 2025 Research Scientist Intern — LightSpeed Studios
High-resolution 3D generation MLLM with a two-stage policy: text-to-saliency prediction plus energy-based token rearrangement, compressing sequence length and inference cost. Supervised by Prof. Xifeng Gao.
Jan – May 2025 Research Assistant — UT Dallas
Robustness pipeline for Vision-Language-Action models under corner cases (lighting shifts, occlusions): a DINO-based out-of-distribution detector plus prompt-based routing that adapts policies or triggers fallbacks — improving grasping reliability without retraining.
Apr – Nov 2023 Machine Learning Engineer Intern — JD.com
Heterogeneous GNNs to improve customer conversion rates in recommendation systems. Supervised by Prof. Desheng Zhang.
2020 – 2023 Machine Learning Engineer Intern — i-Align
High-resolution 3D processing: meshes with millions of faces (CGF) and volume data beyond 2048³ (JOV). Supervised by Prof. Yubo Tao.

Education

2023 – 2028 PhD in Computer Science — University of Texas at Dallas
Geometric deep learning and 3D generation, advised by Prof. Xiaohu Guo. First-author publications at CVPR (Highlight), ECCV and AAAI.
2020 – 2023 MSc in Computer Science — Zhejiang University, CAD&CG State Key Lab
High-resolution 3D data processing, advised by Prof. Yubo Tao and Prof. Hai Lin. Publications in CGF and IJCAI.
2016 – 2020 BSc in Geographic Information System — Central South University
Ranked 2/58. Ricci flow for geometric problems in GIS, advised by Prof. Haifeng Li.

Awards

2020 Outstanding Graduate Student of Hunan Province
2017 – 2019 Seagate Scholarship
2017 Excellent Prize, "ESRI Cup" GIS Software Development Competition