Wei Yao

Hi there, welcome! I am currently a fourth-year Ph.D. student at the Gaoling School of Artificial Intelligence, Renmin University of China. I am truly honored to be advised by Prof. Yong Liu. Now I’m a visiting student at the National University of Singapore, collaborating with Prof. Yunbei Xu. I am expected to graduate in 2027 and am open to research-oriented opportunities in academia and industry, including postdoctoral and research scientist positions.

From October 2023 to March 2024, as a research intern at Shanghai AI Laboratory, I was fortunate to work under the guidance of Prof. Jing Shao. Prior to my Ph.D. studies, I earned my Bachelor of Engineering in Software Engineering from Huazhong University of Science and Technology in June 2022. I was fortunate to be advised by Prof. Kun He. During my undergraduate studies, I was honored to receive the National Scholarship (2019), a recognition that motivated me to pursue further research in AI.

Research Interests

I believe that as AI systems continue to become more capable, a central challenge in the coming years will be how to supervise AI without ground truth. Motivated by this perspective, my research focuses on how LLMs can learn under imperfect supervision, including weak supervision, weak-to-strong generalization, and preference-based alignment. I aim to study these questions across different stages of model development, from pre-training to fine-tuning, and to provide insights from both theoretical analysis and empirical validation.

My recent work studies how strong models can learn reliably from weak, noisy, or limited supervision, as demonstrated by our work in ICML 2026, ACL25 and several upcoming preprints (such as arXiv:2605.05710). My previous research focused on trustworthy AI, including fairness, robustness and interpretability, with publications at ICML25, TMLR24, ACL24, CVPR23.

Preprints

(* indicates equal contribution, # indicates corresponding authors)

On the Blessing of Pre-training in Weak-to-Strong Generalization

Wei Yao, Wang Zhaoyang, Gengze Xu, Chen Qian, Dongrui Liu, Ziqiao Wang, Yong Liu#, Yunbei Xu#
arXiv preprint arXiv:2605.05710

On Weak-to-Strong Generalization and f-Divergence

Wei Yao*, Gengze Xu*, Huayi Tang, Wenkai Yang, Donglin Di, Ziqiao Wang, Yong Liu#
arXiv preprint arXiv:2506.03109

Selected Publications

(* indicates equal contribution, # indicates corresponding authors)

Service

Reviewer: ICML, NeurIPS, ICLR, ACL, CVPR.

Wei Yao

Wei Yao

Research Interests

Preprints

On the Blessing of Pre-training in Weak-to-Strong Generalization

On Weak-to-Strong Generalization and f-Divergence

Selected Publications

On the Emergence of Weak-to-Strong Generalization: A Bias-Variance Perspective

Revisiting Weak-to-Strong Generalization in Theory and Practice: Reverse KL vs. Forward KL

Understanding Model Ensemble in Transferable Adversarial Attack

Understanding Fairness Surrogate Functions in Algorithmic Fairness

Towards Tracing Trustworthiness Dynamics: Revisiting Pre-training Period of Large Language Models

Fair Scratch Tickets: Finding Fair Sparse Networks without Weight Training

Service