PeopleofSuperintelligence

Researcher

Jian Cheng

Also known as 程健

Jian Cheng is a professor at the Institute of Automation, Chinese Academy of Sciences, whose research on neural-network quantization and pruning reduces memory use and speeds up model inference. After earning mathematics degrees at Wuhan University in 1998 and 2001 and a doctorate in pattern recognition at the Institute of Automation in 2004, he worked as a postdoctoral researcher at Nokia Research Center from 2004 to 2006 before joining the National Lab of Pattern Recognition. In computer vision and model compression, he coauthored cascade hashing for 3D reconstruction in 2014, Bit-Split and Stitching for post-training quantization in 2020 and 2023, and constrained optimization for mixed-precision quantization in 2021. He coauthored EAC-MoE in 2025 to compress mixture-of-experts language models, and at the Institute of Automation he studies deep learning, computer vision, and AI-chip design.

Timeline

  1. 1998–2004

    Mathematics and a doctorate

    He earns bachelor's and master's degrees in mathematics from Wuhan University in 1998 and 2001, then completes a doctorate in pattern recognition and intelligent systems at the Institute of Automation, Chinese Academy of Sciences, in 2004.

  2. 2004–2006

    Postdoc at Nokia Research Center

    From 2004 to 2006, he works as a postdoctoral researcher at Nokia Research Center, then joins the National Lab of Pattern Recognition at the Institute of Automation.

  3. 2014

    Image matching for 3D reconstruction

    In 2014, with Cong Leng, Jiaxiang Wu, Hainan Cui, and Hanqing Lu, he coauthors a CVPR paper on fast and accurate image matching with cascade hashing for 3D reconstruction.

  4. 2020

    Bit-split post-training quantization

    He coauthors Bit-Split and Stitching in 2020, a post-training network quantization method that quantizes FP32 models to INT3 with near-original model performance and without fine-tuning.

  5. 2021

    Mixed-precision network quantization

    An ICCV 2021 paper he coauthors formulates mixed-precision neural-network quantization as constrained optimization, reformulates it as a Multiple Choice Knapsack Problem, and solves it with a greedy search algorithm.

  6. 2023

    Optimization-based post-training quantization

    With Peisong Wang, Weihan Chen, Xiangyu He, Qiang Chen, and Qingshan Liu, he coauthors Optimization-based Post-training Quantization with Bit-split and Stitching in IEEE Transactions on Pattern Analysis and Machine Intelligence in 2023.

  7. 2025

    Compressing mixture-of-experts models

    In 2025, he coauthors EAC-MoE, an expert-selection aware compressor for mixture-of-experts language models that calibrates expert routing during quantization and prunes less-used experts to reduce memory usage and improve inference speed.

  8. As of October 2026

    Professor at the Institute of Automation

    As of October 2026, he is a professor at the Institute of Automation, Chinese Academy of Sciences, where he studies deep learning, computer vision, and AI-chip design.

Career

  1. NowChinese Academy of SciencesProfessor

Education

  1. 2004Chinese Academy of SciencesDoctorate in pattern recognition and intelligent systems
  2. 2001Wuhan UniversityMaster's degree in mathematics
  3. 1998Wuhan UniversityBachelor's degree in mathematics

From the biography.

  1. Universities keep AI researchers for a median of 6 years, compared with 2 years at frontier labsIncludes Chinese Academy of Sciences. Across 15,811 dated work stays (6,529 people), median role tenure runs from 6 years at universities to 2 years at frontier labs. We compare all five organization types, starting decades from the 1970s to the 2020s, and the organizations that keep people longest.
  2. People shaping AI work at a median of 3 organizations over a career: 3 for university starters and institute and lab starters, versus 2 for company startersIncludes Chinese Academy of Sciences. Across 6,853 people shaping AI with a dated work role, we estimate a median of 3 parent organizations over a career (mean 3.6), comparing role groups, starting sectors and starting decades from the 1950s to 2020–2026.
  3. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes Chinese Academy of Sciences. We estimate where the people shaping AI worked when their most-cited papers appeared: the company share was 31% in 2020–2024 and 8% in 1970–1974, and Google led all organizations in 2020–2024.
  4. Google has the highest combined h-index among AI organizationsIncludes Chinese Academy of Sciences. We estimate that Google holds the highest combined h-index at 8,016, while universities account for 73% of the deduplicated total across AI organizations.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Jian Cheng yet. Send us a free-licensed photo.

  1. 程健-中国科学院大学-UCAS people.ucas.ac.cn. Checked 22 October 2026.
  2. Jian Cheng's homepage nlpr.ia.ac.cn. Checked 22 October 2026.
  3. Towards Accurate Post-training Network Quantization via Bit-Split and Stitching proceedings.mlr.press. Checked 22 October 2026.
  4. Towards Mixed-Precision Quantization of Neural Networks via Constrained Optimization openaccess.thecvf.com. Checked 22 October 2026.
  5. 程健----中国科学院自动化研究所 ia.cas.cn. Checked 22 October 2026.
  6. EAC-MoE: Expert-Selection Aware Compressor for Mixture-of-Experts Large Language Models aclanthology.org. Checked 22 October 2026.