I am a Research Engineer working on foundation models at 
Kling AI, 
Kuaishou Technology. I received my Ph.D. in Fluid Mechanics from 
Peking University in 2020 under the supervision of Prof. Yue Yang, and was named an Outstanding Graduate. Admitted to Peking University through the Chinese Physics Olympiad (CPhO), I earned B.S. degrees in Mechanics and Economics in 2015.
I joined Kuaishou through its KuaiStar Top Talent Program in 2020 and have received top performance ratings for five consecutive years. I have end-to-end experience training hundred-billion-parameter foundation models and have published 15+ papers at leading venues, including ICML, CVPR, ECCV, and AAAI. My research interests include agentic coding LLMs, video generation models, and Physical AI and embodied intelligence, building on my earlier work in image/video processing, restoration, and enhancement.
现任快手可灵 AI 大模型算法工程师。本科（CPhO 物理竞赛保送）和博士（优秀毕业生）毕业于北京大学，2020 年通过快Star顶尖人才计划加入快手，连续五年取得 Top 绩效。具备千亿参数大模型全链路训练经验，在 ICML、CVPR、ECCV、AAAI 等顶级会议发表论文 15+ 篇。目前主要关注 Agentic Coding 大语言模型、视频生成模型，以及 Physical AI 与具身智能，此前研究涵盖图像/视频处理与增强。
News
- 2026.07: KAT-Coder-Pro V2.5, a trillion-parameter-scale coding LLM with performance approaching GLM-5.2 and Claude Opus 4.8, was released. Its lightweight variant, KAT-Coder-V2.5-Dev, was open-sourced at the same time; the broader model ecosystem has accumulated over 400K downloads and drawn strongly positive feedback from the community.
- 2026.06: One paper accepted to ECCV 2026.
- 2026.05: One paper accepted to ICML 2026.
- 2026.03: KAT-Coder-Pro V2 released, ranked Top 10 on the Artificial Analysis Coding Index.
- 2026.02: Two papers accepted to CVPR 2026, one selected as Highlight.
- 2025.05: One paper accepted to ICML 2025.
- 2025.01: One paper accepted to AAAI 2025.
- 2024.07: Two papers accepted to ECCV 2024.
- 2024.02: Two papers accepted to CVPR 2024.
Models
KAT-Coder-V2.5-Dev
A lightweight 35B-A3B MoE variant of KAT-Coder-V2.5, built for long-horizon agentic coding.
- Agentic coding: 69.4% on SWE-bench Verified and 93.4% on PinchBench.
- Long-context: native support for up to 1M tokens for longer-context serving.
- Reliable tool use: fewer malformed tool labels and repetitive single-turn behaviors.
LPM: Industrial-Scale Generative Video Restoration
LPM Team, Kuaishou Technology
- Large Processing Model (LPM) is a diffusion-based generative framework for photorealistic video restoration
- To our knowledge, LPM is the first generative video restoration model deployed at industrial scale
Publications
* denotes corresponding author; † denotes project leader.
No Time Like the Present: Agentic Test-Time Training for LLM Agents
Yanbo Wang, Jinhua Hao*, Yuze Shi, Kun Yuan, Ming Sun
arXiv 2026 Paper
MMG2Skill: Can Agents Distill In-the-Wild Guides into Self-Evolving Skills?
Xinyu Che, Junqi Xiong, Yunfei Ge, Xinping Lei, Shihao Li, Hang Yan, Han Li, Yuanxing Zhang, Zhiqi Bai, Jinhua Hao, Ming Sun, Han Li, Jiaheng Liu
arXiv 2026 Paper
ICML 2026Coloring the Noise: Adversarial Sobolev Alignment for Faithful Image Super Resolution, Hongbo Wang, Huaibo Huang, Pin Wang, Jinhua Hao†, Chao Zhou, Ran He CodeECCV 2026Tuning Real-World Image Restoration at Inference: A Test-Time Scaling Paradigm for Flow Matching Models, Purui Bai, Junxian Duan, Pin Wang, Jinhua Hao†, Ming Sun, Chao Zhou, Huaibo HuangCVPR 2026ShiftLUT: Spatial Shift Enhanced Look-Up Tables for Efficient Image Restoration, Xiaolong Zeng, Yitong Yu, Shiyao Xiong, Jinhua Hao, Ming Sun, Chao Zhou, Bin Wang CodeCVPR 2026 HighlightInstantViR: Real-Time Video Inverse Problem Solver with Distilled Diffusion Prior, Weimin Bai, Suzhe Xu, Yiwei Ren, Jinhua Hao, Ming Sun, Wenzheng Chen, He Sun ProjectICML 2025Visual Autoregressive Modeling for Image Super-Resolution, Yunpeng Qu, Kun Yuan, Jinhua Hao, Kai Zhao, Qizhi Xie, Ming Sun, Chao Zhou Code PostAAAI 2025Plug-and-Play Tri-Branch Invertible Block for Image Rescaling, Jingwei Bao, Jinhua Hao*†, Pengcheng Xu, Ming Sun, Chao Zhou, Shuyuan Zhu Code PostECCV 2024OAPT: Offset-Aware Partition Transformer for Double JPEG Artifacts Removal, Qiao Mo, Yukang Ding, Jinhua Hao*†, Qiang Zhu, Ming Sun, Chao Zhou, Feiyu Chen, Shuyuan Zhu CodeCVPR 2024CPGA: Coding Priors-Guided Aggregation Network for Compressed Video Quality Enhancement, Qiang Zhu, Jinhua Hao†, Yukang Ding, Yu Liu, Qiao Mo, Ming Sun, Chao Zhou, Shuyuan Zhu Code PostECCV 2024XPSR: Cross-modal Priors for Diffusion-based Image Super-Resolution, Yunpeng Qu, Kun Yuan, Kai Zhao, Qizhi Xie, Jinhua Hao, Ming Sun, Chao Zhou Code PostCVPR 2024PTM-VQA: Efficient Video Quality Assessment Leveraging Diverse PreTrained Models from the Wild, Kun Yuan, Hongbo Liu, Mading Li, Muyi Sun, Ming Sun, Jiachao Gong, Jinhua Hao, Chao Zhou, Yansong TangarXiv 2024CasSR: Activating Image Power for Real-World Image Super-Resolution, Haolan Chen, Jinhua Hao†, Kai Zhao, Kun Yuan, Ming Sun, Chao Zhou, Wei HuJ. Fluid Mech.Effects of magnetic–vortical interactions on magnetic splitting, Linlin Kang, Yanru Wang, Zhongzheng Jiang, Jinhua Hao, Shiying Xiong, Dixia Fan, Weicheng CuiJ. Fluid Mech.Magnetic knot cascade via the stepwise reconnection of helical flux tubes, Jinhua Hao, Yue Yang PKU NewsJ. Fluid Mech.Tracking vortex surfaces frozen in the virtual velocity in non-ideal flows, Jinhua Hao, Shiying Xiong, Yue Yang PKU News
Competitions
CVPR Workshop 2024🏆 2nd Place — NTIRE 2024 Challenge on Image Super-Resolution (×4): Methods and Results, Jinhua Hao*, Kai Zhao, Kun Yuan, Ming Sun, Chao Zhou AwardCVPR Workshop 2024🏆 5th Place — NTIRE 2024 Challenge on Blind Enhancement of Compressed Image: Methods and Results, Qiao Mo, Jingwei Bao, Jinhua Hao*, Yukang Ding, Hantang Li, Ming Sun, Chao Zhou, Shuyuan Zhu
Projects
- 2026.07: 研发并上线代码大语言模型 KAT-Coder-V2.5，性能接近 GLM-5.2 和 Claude Opus 4.8；同步开源轻量版 KAT-Coder-V2.5-Dev，主模型及其衍生版本累计下载量突破 40 万，获得了积极的社区反馈。Developed and deployed KAT-Coder-V2.5, a coding LLM with performance approaching GLM-5.2 and Claude Opus 4.8, and open-sourced its lightweight variant, KAT-Coder-V2.5-Dev. Cumulative downloads across the open-weight model and its derivatives surpassed 400K, and the release received positive community feedback. Hugging Face News
- 2026.03: 研发并上线大语言模型 KAT-Coder V2，在 Artificial Analysis Coding Index 中位列全球第六、国产模型第一（截至 2026 年 3 月 31 日）。Developed and deployed KAT-Coder V2, an agentic coding LLM that ranked sixth globally and first among Chinese models on the Artificial Analysis Coding Index (as of March 31, 2026). News
- 2025.08: 研发并上线生成式视频处理大模型 LPM，在保持保真度的同时提升视频清晰度并生成纹理细节，取得带宽效率与用户观看时长的双重收益。Developed and deployed LPM, a diffusion-based generative video processing model that improves video clarity and generates texture detail while preserving fidelity, delivering gains in both bandwidth efficiency and user watch time. News
- 2023.01: 研发并上线服务端预处理算法，将低质视频智能重塑为高清输出，提升用户观看体验和整体 QoE。Developed and deployed server-side preprocessing algorithms that intelligently restore low-quality videos to high-definition outputs, improving the viewing experience and overall QoE. News
- 2022.07: 研发并上线移动端后处理算法，提升用户播放视频清晰度；面向 NPU/GPU 芯片优化算法，兼顾视觉效果与功耗。Developed and deployed on-device post-processing algorithms optimized for mobile NPU/GPU hardware, improving playback clarity while balancing visual quality and power efficiency. News
- 2021.12: 研发并上线视频质量评估算法，实现对视频画质的多维度感知、理解与评估。Developed and deployed video quality assessment algorithms that analyze video quality across multiple perceptual dimensions. News
Honors
- 2024, 2026: 快手优秀项目奖（2 次）Kuaishou Outstanding Project Award (twice)
- 2023, 2024, 2025: 快手音视频个人荣誉奖（3 次）Kuaishou Audio and Video Outstanding Individual Award (three times)
- 2023: 快手洛子峰奖（2 次）Kuaishou Lhotse Award (twice)
- 2022, 2023: 快手乔戈里奖（2 次）Kuaishou Chogori Award (twice)
- 2020: 快手“快Star”特别技术人才计划（Top 1%）Kuaishou “KuaiStar” Special Technical Talent Program (Top 1%)
- 2020: 北京大学优秀毕业生（Top 1%）Peking University Outstanding Graduate (Top 1%)
- 2015, 2016, 2017, 2019: 北京大学校长奖学金（Top 1%）Peking University President’s Scholarship (Top 1%)
Education
- 2015.09 - 2020.06, Ph.D. in Fluid Mechanics, Peking University, Beijing.
- 2011.09 - 2015.06, B.S. in Mechanics, Peking University, Beijing.