Daiwei Chen (陈戴纬)

I'm a second year Electrical and Computer Engineering (ECE) PhD in Machine Learning and Optimization Theory (MLOPT) Research Group at the University of Wisconsin-Madison.

Under the guidance of my advisor, Prof. Ramya Korlakai Vinayak, my research focuses on

🛠️ LLM Post-Training ( AI Alignment and RLHF )
🕵🏻‍♂️ Mechanistic Interpretability of LLM

Before starting my PhD journey, I obtained my Master's degree in Systems Engineering from the University of Pennsylvania under the guidance of Prof. Pratik Chaudhari, and received my Bachelor of Science in Psychology from Zhejiang University.

news

Jan 22, 2025	We are thrilled to announce that our paper, PAL: Sample-Efficient Personalized Reward Modeling for Pluralistic Alignment, has been accepted to ICLR 2025! - Wish to see you at the Singapore EXPO from April 24 to April 28. 🤗
Dec 08, 2024	Thrilled to attend NeurIPS 2024 in Vancouver next week! Excited to connect with brilliant researchers and exchange ideas. Honored that our recent research on Pluralistic Alignment has been accepted for a 🔥 Spotlight Talk at the NeurIPS 2024 Behavioral ML workshop (also been accepted by NeurIPS AFM / FITML / Pluralistic Alignment / SoLaR workshop 2024)! — wish to see you there! 🤗
Feb 16, 2024	Our blogpost Unraveling The Impact of Training Samples has been accepted by ICLR 2024!

selected publications

PAL: Sample-Efficient Personalized Reward Modeling for Pluralistic Alignment

Daiwei Chen, Yi Chen, Aniket Rege, Zhi Wang, and Ramya Korlakai Vinayak

In The Thirteenth International Conference on Learning Representations, 2025

Abs arXiv Bib Blog

Foundation models trained on internet-scale data benefit from extensive alignment to human preferences before deployment. However, existing methods typically assume a homogeneous preference shared by all individuals, overlooking the diversity inherent in human values. In this work, we propose a general reward modeling framework for pluralistic alignment (PAL), which incorporates diverse preferences from the ground up. PAL has a modular design that leverages commonalities across users while catering to individual personalization, enabling efficient few-shot localization of preferences for new users. Extensive empirical evaluation demonstrates that PAL matches or outperforms state-of-the-art methods on both text-to-text and text-to-image tasks: on Reddit TL;DR Summary, PAL is 1.7% more accurate for seen users and 36% more accurate for unseen users compared to the previous best method, with 100× less parameters. On Pick-a-Pic v2, PAL is 2.5% more accurate than the best method with 156× fewer learned parameters. Finally, we provide theoretical analysis for generalization of rewards learned via PAL framework showcasing the reduction in number of samples needed per user.
@inproceedings{chen2025pal, url = {https://openreview.net/forum?id=1kFDrYCuSu}, title = {<a href="https://openreview.net/forum?id=1kFDrYCuSu" target="_blank">{PAL}: Sample-Efficient Personalized Reward Modeling for Pluralistic Alignment</a>}, booktitle = {The Thirteenth International Conference on Learning Representations}, author = {Chen, Daiwei and Chen, Yi and Rege, Aniket and Wang, Zhi and Vinayak, Ramya Korlakai}, year = {2025}, eprint = {2406.08469}, archiveprefix = {arXiv}, primaryclass = {id='cs.LG' full_name='Machine Learning' is_active=True alt_name=None in_archive='cs' is_general=False description='Papers on all aspects of machine learning research (supervised, unsupervised, reinforcement learning, bandit problems, and so on) including also robustness, explanation, fairness, and methodology. cs.LG is also an appropriate primary category for applications of machine learning methods.'} }