Low-Precision Computing at CNCC: New Frontiers in Data Formats, Algorithms, and System Co-Design
CNCC2026Forum
In the era of large models, how can fewer bits deliver greater computational efficiency? On October 22, the GCC-HiFloat community will host the Low-Precision Computing Forum at CNCC 2026. The forum will focus on frontier research in low-precision data formats, model training and inference, quantization algorithms, hardware–software co-design, and system optimization, examining key challenges and recent advances from numerical representation to efficient AI computing systems.
Low precision involves more than simply removing a few bits. As bit widths decrease, trade-offs among model accuracy, numerical dynamic range, quantization error, training stability, and hardware execution efficiency become more pronounced. Many questions remain: how to design new low-precision formats that better match AI data distributions; how to coordinate training, inference, and quantization algorithms; and how compilation, operators, and architecture can translate theoretical bit-width advantages into end-to-end performance gains.
The forum will explore the academic frontiers of low-precision computing, covering numerical representations and data formats, low-bit model training and inference, quantization algorithms, hardware–software co-design, efficient operators and architectures, and low-precision computing systems and data transmission. Experts from relevant fields in China will share their latest findings and discuss foundational questions, key technologies, and future research trends.
Forum Agenda:
| No. | Topic | Speaker | Organization |
|---|---|---|---|
| 1 | Low-Precision Operator Development and Performance Optimization | Lu Lu | South China University of Technology |
| 2 | Numerical Precision in Large Models: Misconceptions, Principles, and Bottlenecks | Li Shang | Fudan University |
| 3 | AI Computing System Design and Optimization | Peisong Wang | Institute of Automation, Chinese Academy of Sciences |
| 4 | Hardware–Software Co-Optimization of Low-Precision Data Formats for AI Computing | Jinming Lu | Nanjing University |
| 5 | Ultra-Low-Bit Large-Model Inference: From Numerical Representation to Efficient Execution | Jingwen Leng | Shanghai Jiao Tong University |
| 6 | Toward Foundation Models at Extremely Low Precision: From Numerical Values and Tensors to Temporal Reuse | Haotong Qin | The Chinese University of Hong Kong |
| 7 | Introduction to HiFloat Formats and Practical Training and Inference | Sitan Li | Huawei Technologies Co., Ltd. |
Time: October 22, 13:30–17:30
Venue: Room 107A, Sanxingdui A, Chengdu Tianfu International Convention Center
Contact: guizelin@gccorg.com