Jiaming Zhang

Jiaming Zhang

Undergraduate @RUC

About me

I am a first-year master’s student at the School of Statistics, Renmin University of China.
I am currently seeking PhD opportunities for Fall 2027.

Currently, I am working as a remote intern at the Provable Responsible AI and Data Analytics (PRADA) Lab at KAUST, advised by Prof. Di Wang.

Also, I am collaborating with the TEA Lab at the University of British Columbia (UBC).

Research Interests

My general research interest lies in Trustworthy AI. More specifically, I am currently focusing on the following directions:

  • Privacy in Reinforcement Learning with Human Feedback (RLHF):
    Exploring how to protect user/annotator privacy in the training process of large language models guided by human feedback.

  • Benign Overfitting in Large Language Models (LLMs):
    Studying the phenomenon where overparameterized models generalize well despite fitting noisy or mislabeled data.
    This includes analyzing the learning dynamics of LLMs to better understand how and why benign overfitting occurs.

  • Data Valuation/Attribution:
    Developing efficient methods for estimating the value and influence of individual training data points, with applications to data pruning and selection.

I am broadly interested in making machine learning systems more transparent, reliable, and accountable.

News

Publications

* indicates co-first authors.

  • Towards User-level Private Reinforcement Learning with Human Feedback [arxiv]

    Jiaming Zhang*, Mingxi Lei*, Meng Ding, Mengdi Li, Zihang Xiang, Difei Xu, Jinhui Xu, Di Wang

    Conference on Language Modeling (CoLM 2025)

  • Efficient Forward-Only Data Valuation for Pretrained LLMs and VLMs [arxiv]

    Wenlong Deng, Jiaming Zhang, Qi Zeng, Christos Thrampoulidis, Boying Gong, Xiaoxiao Li

    Submitted to AAAI 2026

Teaching

  • Teaching Assistant of Regression Analysis, 2024 fall, Renmin University of China