CUHK
academicThe Chinese University of Hong Kong (est. 1963, Sha Tin). This entry deliberately groups CUHK with CUHK-Shenzhen, its 2014-founded, legally distinct sister campus — because the highest-velocity artifact producer in the CUHK orbit is CUHK-Shenzhen's FreedomIntelligence group (Benyou Wang, School of Data Science): the HuatuoGPT-o1/-3 medical-reasoning line, the TinyDeepSeek from-scratch DeepSeek-R1 reproduction, the ShizhenGPT TCM foundation model, the AceGPT Arabic series, and the TwinMarket financial-agent simulator. Shenzhen-campus outputs are flagged (CUHK-SZ) in their descriptions.
On the Hong Kong campus, Hongsheng Li's MMLab-CUHK is the most productive LLM+multimodal group: MathCoder2/MathCode-Pile (widely reused 19.2B-token math pretraining corpus), the MATH-Vision benchmark, and WebGen-Bench; CameraCtrl II is heavy ByteDance co-development. Michael Lyu's ARISE lab runs the software-engineering evaluation line (SWE-Chain, CodeCrash), and Pheng-Ann Heng leads medical-imaging and surgical AI (Surg-Σ ecosystem with Qi Dou and NVIDIA).
Attribution flags matter here more than anywhere: Dahua Lin's large-model output (InternLM, OpenMMLab, Lumina) runs through Shanghai AI Lab and files there; Jiaya Jia moved to HKUST, taking the dvlab line with him; and Huawei openPangu checkpoints hosted on the FreedomIntelligence HuggingFace org file under Huawei.
People
- Hongsheng Li Google Scholar — Professor, EE; MMLab-CUHK & CPII (MathCoder, WebGen-Bench, CameraCtrl II)
- Dahua Lin Google Scholar — Professor, IE; MMLab co-lineage — large-model output ships via Shanghai AI Lab
- Pheng-Ann Heng Google Scholar — Chair Professor, CSE; medical imaging & surgical AI; Institute of Medical Intelligence and XR
- Michael R. Lyu Google Scholar — Chair Professor, CSE; ARISE lab (SWE-Chain, CodeCrash)
- Benyou Wang Google Scholar — Associate Professor, CUHK-Shenzhen SDS; leads FreedomIntelligence (HuatuoGPT, TinyDeepSeek, AceGPT)