Cong Fang

Assistant Professor, Peking University

Joined

January 2019

Names

Cong Fang (Preferred)

Emails

****@pku.edu.cn (Confirmed)

Personal Links

Homepage

Google Scholar

DBLP

Career & Education History

Assistant Professor

Peking University (pku.edu.cn)

2021 – Present

PhD student

Peking University (pku.edu.cn)

2014 – 2019

Advisors, Relations & Conflicts

No relations added

Expertise

No areas of expertise listed

Publications

Scaling Law for SGD in Quadratically Parameterized Linear Regression
Shihong Ding, Haihan Zhang, Hanzhen Zhao, Cong Fang
- ICLR 2026 Conference Withdrawn Submission
- Readers: Everyone
AdaPM: a Partial Momentum Algorithm for LLM Training
Yimu Zhang, Yuanshi Liu, Cong Fang
- Submitted to ICLR 2026
- Readers: Everyone
Preference-Based Process Reward Model for Robust Mathematical Reasoning
Yimu Zhang, Cong Fang
- Submitted to ICLR 2026
- Readers: Everyone
Strong Correlations Induce Cause Only Predictions in Transformer Training
Haihan Zhang, Yimu Zhang, Cong Fang
- ICLR 2026 Poster
- Readers: Everyone
Generalization Analysis of SGD in Linear Regression under Covariate Shift: A View from Preconditioning
Yuanshi Liu, Haihan Zhang, Qian Chen, Cong Fang
- ICLR 2026 Conference Withdrawn Submission
- Readers: Everyone
Implicit bias of Hessian Approximation in Regularized Randomized SR1 Method
Kecen Sha, Yuanshi Liu, Weicheng Lin, Cong Fang
- Submitted to ICLR 2026
- Readers: Everyone
MVR: Multi-view Video Reward Shaping for Reinforcement Learning
Lirui Luo, Guoxi Zhang, Hongming Xu, Yaodong Yang, Cong Fang, Qing Li
- ICLR 2026 Poster
- Readers: Everyone
Conda: Column-Normalized Adam for Training Large Language Models Faster
Junjie Wang, Pan Zhou, Yiming Dong, Huan Li, Jia Li, zhou Xun, Qicheng Lao, Cong Fang, Zhouchen Lin
- ICLR 2026 Conference Withdrawn Submission
- Readers: Everyone
PaZO: Preconditioned Accelerated Zeroth-Order Optimization for Fine-Tuning LLMs
Hanzhen Zhao, Shihong Ding, Cong Fang, Zhouchen Lin
- NeurIPS 2025 poster
- Readers: Everyone
Improving Model Representation and Reducing KV Cache via Skip Connections with First Value Heads
Zhoutong Wu, Yuan Zhang, Yiming Dong, Chenheng Zhang, Cong Fang, Kun Yuan, Zhouchen Lin
- NeurIPS 2025 poster
- Readers: Everyone

View all 57 publications

Co-Authors

View all 71 co-authors