Education
Bachelor of Engineering
GPA 3.98 / 4.00. Ranked #1 out of 204 in the Department of CS.
Proud to have received the highest distinction for undergraduates at Tsinghua in 2023 and the SenseTime Scholarship in 2024.
Ph.D. Student
Advisor: Prof. Ya-Qin Zhang, Dean of Institute for AI Industry Research (AIR), THU.
Research Interests
I am a firm believer in The Bitter Lesson: throughout the history of AI, general methods that can leverage increasing amounts of computation have ultimately outperformed approaches built around hand-crafted knowledge. Recent advances in large-scale reinforcement learning for LLM reasoning and agentic coding have further strengthened this belief. They suggest that scaling learning through exploration, interaction, and verifiable feedback is a promising path toward increasingly capable intelligence.
Looking ahead, I believe autonomous research systems will develop along two dimensions: depth and breadth. In depth, these systems will push the limits of LLM intelligence by scaling the number of tokens devoted to solving open problems. In breadth, they will expand into real-world R&D workflows. In large technology companies, for example, an AI researcher could work alongside algorithm engineers: given an objective, a concrete key result, and a baseline, it would continuously explore different approaches and improve the target without reward hacking.
My research aims to help realize this vision by developing the data, benchmarks, and frontier learning algorithms needed to build and evaluate such systems systematically. Two questions are central to this agenda. First, how can we train these systems effectively over long horizons, where useful feedback may be sparse and credit must be assigned across extended sequences of decisions and experiments. Second, how can we encode human judgment and taste into the training signal, so that these systems learn not only to optimize measurable outcomes, but also to recognize which ideas are promising, meaningful, and worth pursuing.
If you’re interested in related topics and would like to collaborate, feel free to reach out! You can find my email in the hyperlink on the right panel.
Past Research Experience
Generative Simulation for Embodied AI
Problem Identification
The development and iteration of autonomous driving and robotics policies are limited by the high costs, low efficiency, and safety risks of real-world testing. The rise of Generative AI offers a potential breakthrough by enabling high-fidelity, interactive, and editable simulation testing.
Technical Approach
My research focused on building "World Models" to drive simulation with generative methods. Technically, I explored two core directions: 1) High-fidelity Scene Reconstruction: Building "digital twins" of real scenes using technologies like NeRF or Gaussian splatting. 2) Controllable Content Generation: On the basis of reconstructed scenes, leveraging the generative priors of diffusion models to provide endless, controllable scene variations and edge cases.
Selected Publications
- †PartRM: Modeling Part-Level Dynamics with Large Cross-State Reconstruction Model, CVPR 2025.
- †Ctrl-U: Robust Conditional Image Generation Via Uncertainty-aware Reward Modeling, ICLR 2025.
- †SCP-Diff: Spatial-Categorical Joint Prior for Diffusion Based Semantic Image Synthesis, ECCV 2024.
Data Efficient Scene Parsing
Problem Identification
2D/3D perception is fundamental to embodied intelligence, but the extremely high cost of data annotation severely restricts the development of perception models.
Technical Approach
My early research focused on data-efficient perception learning algorithms, particularly semi-supervised learning and domain adaptation. In my first ICCV paper, DQS3D, I proposed a single-stage, densely-matched semi-supervised learning framework for 3D object detection, addressing the issue of insufficient training signals caused by sparse matching in previous methods. I also explored various levels of perception tasks such as self-supervised depth estimation, indoor layout estimation, and HD map generation, mastering task-oriented neural network and representation design methods.
Publications
- †DQS3D: Densely-matched Quantization-aware Semi-supervised 3D Detection, ICCV 2023.
- †From Semi-supervised to Omni-supervised Room Layout Estimation Using Point Clouds, ICRA 2023.
- †Training-Free Model Merging for Multi-target Domain Adaptation, ECCV 2024.
Services
Co-Founder @ Lumina-Embodied.AI
- Building community for embodied AI research and applications
- Bridging academic research with industry implementations
- Focus on AI systems that learn through physical interaction
Reviewer @ Academic Conferences & Journals
- CVPR (2025), ICCV (2025), WACV (2024, 2025), 3DV (2025, 2026), TPAMI
- NeurIPS (2025), ICLR (2025)
- ICRA (2025), IROS (2024, 2025), CoRL (2025)
- AAAI (2024), ICME (2025)
Teaching Assistant @ CS, THU
- (30240163) Software Engineering. Compulsory course in CS, THU. (23Spring, 23Fall, 24Spring, 24Fall, 25Spring, 25Fall)
- (30240551) Digital Logic Experimentation. Compulsory course in CS, THU. (24Spring, 25Spring)
- (40240354) Computer Organization and Design. Compulsory course in CS, THU. (23Fall)
清华大学计算机系 科创辅导员 (2024.9-2026.6)
- Technical training & competition guidance for undergraduates
- Research & internship opportunity integration
清华大学计算机系 学生科协主席 (2023.5-2024.6)
- Built homepage & documentation
- Organized summer training camp for freshmen
- Launched resource portal for students