Haopeng Li

I am an M.Phil. student at HKUST(GZ). I am currently a research intern at NVIDIA Research supervised by Enze Xie. I was previously a visiting student at Westlake University, advised by Prof. Huan Wang. My research focuses on Efficient Generative AI and MLsys.

Projects

Sol-EngineVideo Inference

Sol Video Inference Engine

An agent-native, full-stack framework that accelerates video generation models end to end at NVIDIA.

• Code
• Docs
• Report

1. Jul 2026 Sol-Attn Training-free sparse attention
2. Aug 2026 MiniMax-H3 day-one optimization 3.95× end-to-end on 8× GB200
3. Sep 2026 Sol-H3 5s clip in 1.65 s on 8× B300
4. Oct 2026 Sol-Refiner · Open-H3-Regenerate 2K in 3 steps, 5× faster than the API

Publications Full list on Google Scholar

Sol-Attn: Accelerating Video Generation Inference via On-the-Fly Attention Sparsification

Haopeng Li, Yitong Li, Junsong Chen, Tian Ye, Haozhe Liu, Jincheng Yu, Duomin Wang, Ruihua Zhang, Zeke Xie, Enze Xie, Song Han

• arXiv
• Page
• Code

PISA: Piecewise Sparse Attention Is Wiser for Efficient Diffusion Transformers

Haopeng Li, Shitong Shao, Wenliang Zhong, Zikai Zhou, Lichen Bai, Hui Xiong, Zeke Xie

• NeurIPS 2026
• arXiv
• Code

Autoregressive Image Generation with Randomized Parallel Decoding

Haopeng Li, Jinyue Yang, Guoqi Li, Huan Wang

• ICLR 2026
• arXiv
• Page
• Code

SoL-Refiner: Speed-of-Light One-Step Refinement for High-Resolution Video

Haozhe Liu, Tian Ye, Shuchen Xue, Sensen Gao, Yitong Li, Junsong Chen, Haopeng Li, Jincheng Yu, Duomin Wang, Ruihua Zhang, Lei Zhu, Jia-Wang Bian, Song Han, Enze Xie

• arXiv
• Page
• Code

Sol Video Inference Engine: Agent-Native Full-Stack Acceleration Framework for Efficient Video Generation

Yitong Li, Junsong Chen, Haopeng Li, Haozhe Liu, Jincheng Yu, Ligeng Zhu, Ping Luo, Song Han, Enze Xie

• arXiv
• Docs
• Code

LIVEditor-14B: Lightning Unified Video Editing via In-Context Sparse Attention

Shitong Shao, Zikai Zhou, Haopeng Li, Yingwei Song, Wenliang Zhong, Lichen Bai, Zeke Xie

• ICML 2026
• arXiv
• Code