Highlights of AI Large‑Model Sessions at CNCC 2024

On October 24, the 2024 China Computer Conference (CNCC 2024) opened in Hengdian, Dongyang, hosted by the China Computer Federation.

The first day, Ant Group organized an "AI Large‑Model Technology Exploration and Application" event with experts from Zhejiang University, University of Science and Technology of China, Institute of Computing Technology, Xi'an Jiaotong University, and Renmin University of China, discussing development challenges and solutions.

Professor Chen Enhong (USTC) presented "Large‑Model‑Driven Paradigm Shift in Scientific Research," highlighting applications such as ChemCrow (GPT‑4‑based chemistry tool) and AlphaFold 3 for protein‑structure prediction, and emphasizing the need for robust AI‑for‑Science ecosystems.

Professor Chen Huajun (Zhejiang University) delivered "Knowledge Enhancement and Governance for Large Models," describing how knowledge graphs can augment model knowledge, detect hallucinations, edit outdated or toxic information, and protect copyright and privacy.

Professor Cheng Xueqi (Institute of Computing Technology) spoke on "Data Infrastructure and Broad‑Spectrum Association Analysis," explaining the concept of data spaces, value extraction, and practical association‑analysis techniques with real‑world case studies.

Associate Professor Shi Bin (Xi'an Jiaotong University) introduced "Vertical‑Domain Large‑Model Inference Enhancement," proposing a few‑shot fine‑tuning framework, a pseudo‑labeling method based on factor‑graph modeling, and a bootstrapped chain‑of‑thought approach that boosted zero‑shot performance by up to 20.44 percentage points.

Associate Professor Li Chongxuan (Renmin University) discussed "Frontiers of Diffusion Models," covering continuous diffusion theory, efficient sampling, multimodal generation, 3‑D object creation, and the integration of diffusion models with Transformers.

Senior algorithm expert Chen Jingdong (Ant Group) presented "Ant Multimodal Large Model Progress and Outlook," describing the Baoling multimodal model’s hierarchical feature fusion, visual reasoning chains, and its deployment in over 4 million Alipay mini‑programs and 8 000 life‑service scenarios.

Senior algorithm expert Wang Jian (Ant Group) shared "Alipay Medical Large Model Technology and Applications," detailing challenges in medical AI, the model’s trillion‑scale medical corpus, superior benchmark performance, and its integration into Alipay Health services for diagnosis, policy queries, and personalized health agents.