CV
Academic CV for Jiaxuan Zou.
Contact Information
| Name | Jiaxuan Zou |
| Professional Title | Undergraduate Student in Mathematics and Statistics |
| 3140143497@qq.com | |
| Location | Xi'an, Shaanxi |
| Website | https://jiaxuanzou0714.github.io |
Professional Summary
Undergraduate in Mathematics and Statistics at Xi’an Jiaotong University and research intern in the ByteDance Seed Pre-training team. Previously a research intern at the Gaoling School of Artificial Intelligence, Renmin University of China. My work focuses on mechanistic interpretability, training dynamics, optimizer design, and scaling laws.
Experience
-
2026 - Present Beijing, China
Research Intern
ByteDance, Seed Pre-training Team
-
2025 - 2026 Beijing, China
Research Intern
Gaoling School of Artificial Intelligence, Renmin University of China
Advised by Prof. Yong Liu.
- Work on mechanistic interpretability, deep learning theory, optimizer design, and scaling laws.
- Co-authored preprints on linear attention, neural scaling laws, latent chain-of-thought, and matrix optimizer design.
-
2026 - Present Remote
Co-maintainer
ScalingOpt
A project on optimizer design for large language model training.
- Work on the relation among optimizer design, model architecture, and training configuration under scaling-law regimes.
- Help maintain the optimizer library, including the Nora optimizer entry.
-
Beijing, China
AI Technical Consultant
Tsinghua-affiliated AI startup
- Consult on “AI + K-12 Education” products developed with the Beijing Dongcheng District Education Commission.
-
2024 - Present Xi'an, China
Founder and Organizer
Xi'an Jiaotong University Deep Learning Seminar
- Started an undergraduate deep learning seminar at Xi’an Jiaotong University.
- The seminar later attracted more than one thousand participants from across China and led to research collaborations.
Education
-
2024 - Present Xi'an, China
Undergraduate Student
Xi'an Jiaotong University
Mathematics and Statistics
- School of Mathematics and Statistics.
- Research interests include deep learning theory, optimization, mechanistic interpretability, and scaling laws.
Projects
-
ScalingOpt
A discussion platform and benchmark project for optimizer design in large language model training.
- Co-maintainer.
- Focuses on optimizer design, model architecture, and training configuration under the scaling-law paradigm.
- Nora has been included in the ScalingOpt optimizer library.
Publications
-
2026 -
2026 Nora: Normalized Orthogonal Row Alignment for Scalable Matrix Optimizer
Under Review
Jinghui Yuan, Jiaxuan Zou, Shuo Wang, Yong Liu, and Feiping Nie. arXiv:2605.03769.
-
2026 Effective Frontiers: A Unification of Neural Scaling Laws
Under Review
Jiaxuan Zou, Zixuan Gong, Ye Su, Huayi Tang, and Yong Liu. arXiv:2602.02593.
-
2026 Capabilities and Fundamental Limits of Latent Chain-of-Thought
EMNLP
Jiaxuan Zou, Yaozhong Xiong, and Yong Liu. arXiv:2602.01148.
-
2026 Statistical MIA: Rethinking Membership Inference Attack for Reliable Unlearning Auditing
Under Review
Jialong Sun, Zeming Wei, Jiaxuan Zou, Jiacheng Gong, Guanheng Wang, Chengyang Dong, Jialong Li, and Bo Liu. arXiv:2602.01150.
-
2025 FreIE: Low-Frequency Spectral Bias in Neural Networks for Time-Series Tasks
IEEE ICDM
Jialong Sun, Xinpeng Ling, Jiaxuan Zou, Jiawen Kang, and Kejia Zhang.
-
2026 Mamba-Driven and Feature-Fused U-Net for Automatic Seismic Horizon Interpretation
IEEE Transactions on Geoscience and Remote Sensing
Tian Zhang, Jiaju He, Naihao Liu, Jiaxuan Zou, and Yongxiang Jiang.
-
2024 Seismic Horizon Picking Using Channel-Independent Multi-Scale UNet
Under Review
Jiaxuan Zou, Naihao Liu, Tian Zhang, Jiaju He, Tao Li, and Jinghuai Gao.
Research Interests
- Mechanistic interpretability of LLMs.
- Training dynamics of finite-width neural networks.
- Optimizer design for LLM pre-training.
- Scaling laws and their failure modes.
- Deep learning theory and optimization.