Hongda Liu (Claude)

Hongda Liu (Claude)

Researcher of Vision-Language Models (VLMs)

Welcome to Claude’s Personal Homepage!

Now, I am serving as an Algorithm Researcher at SenseTime Research.

Before that, I received my M.S. degree at Spatial Artificial Intelligence Lab (SAIL), Sun Yat-sen University (SYSU). My advisors are Yulan Guo and Longguang Wang. I received my B.S. degree from the School of Information and Software Engineering, University of Electronic Science and Technology of China (UESTC) in 2023.

My research focuses on efficient Vision-Language Models (VLMs), particularly inference acceleration, model compression, and training optimization. I also have extensive experience in image and video generation and editing, as well as low-level vision and computational photography.

:fire: News

  • 2026.06: :tada::tada: I joined SenseTime Research as an Algorithm Researcher, focusing on inference, compression, and training optimization for VLMs.

  • 2026.04: :tada::tada: One paper (SFLAM) was accepted by TMC 2026.

  • 2026.03: :tada::tada: One paper (POSITION) was accepted by TIP 2026.

  • 2025.07: :tada::tada: I was invited to be a reviewer of International Journal of Computer Vision (IJCV). I am happy to serve as a peer reviewer for conferences/journals/transactions. Feel free to contact me.

  • 2025.02: :tada::tada: One paper (SaMam) was accepted by CVPR 2025 (Highlight, top 3%).

  • 2024.10: :tada::tada: One paper (SaMST) was accepted by ACCV 2024.

  • 2024.03: :tada::tada: As a member of the organizing committee, I organized NTIRE 2024 Stereo Image Super-Resolution Challenge ([Track 1 Constrained SR & Bicubic Degradation] & [Track 2 Constrained SR & Realistic Degradation]). You can read our technical report here.

:briefcase: Work Experience

– Present
SenseTime Research
Algorithm Researcher
Efficient Vision-Language Models (VLMs), with a focus on inference acceleration, model compression, and training optimization.

:book: Education

–
Spatial Artificial Intelligence Lab (SAIL), Sun Yat-sen University (SYSU)
Master of Science (M.S.)
–
University of Electronic Science and Technology of China (UESTC)
Bachelor of Science (B.S.)

:computer: Internship Experience

–
Huawei Cloud
AI Development Engineer
Long-term Memory Service Group (LMS), Chengdu
Large Language Model and Vector Database
–
Tencent
Computer Vision Algorithm Engineer
Platform and Content Group (PCG), Shenzhen
Image/Video Generation and Editing
–
Huawei Cloud
Computer Vision Algorithm Engineer
Physical Intelligence Lab, Shenzhen
3D Scene Alignment and Reconstruction
Advisor: Kaiwen Xue

:books: Selected Publications

SFLAM
Deploying Large AI Models on Resource-Limited Devices with Split Federated Learning
TMC 2026
Xianke Qiang, Hongda Liu, Xinran Zhang, Zheng Chang, Ying-Chang Liang
[paper]
SaMST
POSITION: Open World 3D Scene CAD Recomposition
TIP 2026
Rongkun Yang, Hongda Liu, Yijun Chen, Sheng Ao, Yongjian Zhang, Longguang Wang, Kaiwen Xue, Shunbo Zhou, Yulan Guo
[paper] [code]
SaMam
SaMam: Style-aware State Space Model for Arbitrary Image Style Transfer
CVPR 2025 Highlight
Hongda Liu, Longguang Wang, Ye Zhang, Ziru Yu, Yulan Guo
[paper] [code]
SaMST
Pluggable Style Representation Learning for Multi-Style Transfer
ACCV 2024
Hongda Liu, Longguang Wang, Weijun Guan, Ye Zhang, Yulan Guo
[paper] [code]
ReDSR
Preserving Full Degradation Details for Blind Image Super-Resolution
Arxiv 2024
Hongda Liu, Longguang Wang, Ye Zhang, Kaiwen Xue, Shunbo Zhou, Yulan Guo
[paper] [code]

:wrench: Selected Projects

LightX2V inference framework
LightX2V: Light Image&Video Generation Inference Framework
As a primary developer, I am now focusing on model integration and inference optimization. Recent work includes HunyuanImage-3.0, BAGEL, and SenseNova-Vision.
[code] [HunyuanImage-3.0] [BAGEL] [SenseNova-Vision]
SaMST
PhotoDM: High-Fidelity Photorealistic Style Transfer via Pyramidal High-Frequency Autoencoder and Exact Distribution Matching
[code]