Ximing Xing is a Ph.D. student (2022-Present) in Software Engineering at Beihang University, advised by Professor Qian Yu.
He is dedicated to advancing the frontiers of AI-driven content creation, with a particular focus on the generation and understanding of vector graphics. To this end, his research interests include deep generative models, vector art synthesis, text-to-SVG, neural rendering, SVG diffusion, and SVG LLMs. His research has been published in top-tier international conferences and journals such as CVPR, T-PAMI, and NeurIPS. Some of his representative works include:
- LLMs for Vector Graphics: LLM4SVG[Code]
, Reason-SVG
- Text-to-SVG Synthesis: SVGDreamer++[Code]
, SVGDreamer [Code]
- Vector Diffusion Models: SVGFusion[Code]
- Sketch Synthesis: DiffSketcher[Code]
- Differentiable SVG Rendering Library: PyTorch-SVGRender[Code]
His commitment to open science is reflected in his popular open-source projects, which have collectively garnered a
substantial number of stars on GitHub.
He further contributes to the AI community by sharing models and hosting interactive demonstrations on HuggingFace (@xingxm profile, SVGRender Space).
๐ฅ News
- 2025.02: ย ๐๐ Our paper LLM4SVG has been accepted by CVPR2025!
- 2025.02: ย ๐๐ Our paper SVGDreamer++ has been accepted by T-PAMI ! This is the first paper on SVG generation ever published in T-PAMI.
- 2024.02: ย ๐๐ Our paper SVGDreamer has been accepted by CVPR 2024!
- 2023.12: ย ๐๐ We released PyTorch-SVGRender, a state-of-the-art library for differentiable SVG rendering in PyTorch.
๐ Publications
๐ท๏ธ Topic: Vector Graphics Synthesis
Reason-SVG: Hybrid Reward RL for Aha-Moments in Vector Graphics Generation
Ximing Xing, Yandong Guan, Jing Zhang, Dong Xu, Qian Yu
TL;DR: Reason-SVG introduces the first framework to enhance SVG generation in LLMs through a โDrawing-with-Thoughtโ (DwT) paradigmโcombining explicit design reasoning with codeโtrained via supervised fine-tuning and HyperReward-driven reinforcement learning.
Empowering LLMs to Understand and Generate Complex Vector Graphics
Ximing Xing, Juncheng Hu, Guotao Liang, Jing Zhang, Dong Xu, Qian Yu
TL;DR: LLM4SVG introduces learnable SVG Semantic Tokens and a large SVGX-SFT dataset, enabling LLMs to understand and generate complex vector graphics.
IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2025
SVGDreamer++: Advancing Editability and Diversity in Text-Guided SVG Generation
Ximing Xing, Qian Yu, Chuang Wang, Haitao Zhou, Jing Zhang, Dong Xu
TL;DR: SVGDreamer++ is an advanced text-to-SVG generator with two core innovations: Hierarchical Image Vectorization (HIVE) - enables semantic object-level and component-level image vectorization, and Adaptive Vector Primitive Control โ dynamically assigns the optimal number of vector primitives, capturing fine-grained details without wasting computation.
IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI 2025)
SVGFusion: Scalable Text-to-SVG Generation via Vector Space Diffusion
Ximing Xing, Juncheng Hu, Jing Zhang, Dong Xu, Qian Yu
TL;DR: SVGFusion improves text-to-SVG generation by using a VP-VAE to learn a vector representation of SVG elements, and a VS-DiT to generate SVGs from text prompts by performing diffusion within that learned vector space.
SVGDreamer: Text Guided SVG Generation with Diffusion Model
Ximing Xing, Chuang Wang, Haitao Zhou, Jing Zhang, Dong Xu, Qian Yu
TL;DR: SVGDreamer introduces Semantic-driven Image VEctorization (SIVE) and Vector Particle-based Score Distillation (VPSD) to generate editable, high-quality SVGs with better shape control and diversity.
IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2024
VectorPainter: Advanced Stylized Vector Graphics Synthesis Using Stroke-Style Priors
Juncheng Hu, Ximing Xing, Jing Zhang, Qian Yu
TL;DR: VectorPainter synthesizes text-guided vector graphics by imitating stylized strokes.
DiffSketcher: Text Guided Vector Sketch Synthesis through Latent Diffusion Models
Ximing Xing, Chuang Wang, Haitao Zhou, Jing Zhang, Qian Yu, Dong Xu
TL;DR: DiffSketcher pioneered the use of diffusion models for text-to-vector sketch generation.
Advances in Neural Information Processing Systems (NeurIPS), 2023
๐ท Topic: CAD Generation
CAD-Coder: Text-to-CAD Generation with Chain-of-Thought and Geometric Reward
Yandong Guan, Xilin Wang, Ximing Xing, Jing Zhang, Dong Xu, Qian Yu
TL;DR: CAD-Coder enables LLMs to generate complex, valid 3D CAD models from text by outputting CadQuery (Python) scripts, using a novel two-stage training and chain-of-thought approach.
๐ท Topic: Controllable Text-to-Image Generation
Ximing Xing, Chuang Wang, Haitao Zhou, Zhihao Hu, Chongxuan Li, Dong Xu, Qian Yu
TL;DR: Free training for sketch control image synthesis via Stochastic Differential Equations (SDEs).
๐ท Topic: Robust Machine Learning
A Graph-Based Method for Reasoning About Label Noise
HaiYang Zhang, Ximing Xing, Liang Liu
TL;DR: DualGraph, the first method for label noise processing based on graph neural networks.
IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2021
๐ Projects
Pytorch-SVGRender: A Differentiable Rendering Library for SVG Creation
๐ฅ Main Contributors: Ximing Xing, Juncheng Hu
TL;DR: SVG Differentiable Rendering: Generating vector graphics using neural networks. Support: Text-to-SVG, Image-to-SVG and SVG Editing.
๐ Project | ๐ Code | ๐ค HuggingFace | ๐ Docs
๐ Honors and Awards
- 2025.04 Academic Excellence Foundation of BUAA for PhD Students.
- 2024.12 National Scholarship for Doctoral Students.
- 2021.12 National Scholarship.
๐ Professional Activities
-
Conference Reviewer
CVPR 2024, ECCV 2024, NIPS 2024, ACM MM 2024, AAAI 2025, CVPR 2025, SIGGRAPH 2025, NIPS 2025
-
Journal Reviewer
IJCV, IEEE T-VCG
๐ป Internships
-
2021.06 - 2021.11, Ant Group, machine learning algorithm intern. HangZhou, China.
Multi-Turn Task-Oriented Dialogue System based on deep reinforcement learning.
- Behavior cloning to enhance model cold-start training.
- Agenda-based User Simulator in RL.
- Turns- and sessions-level reward design for PPO training.