Xun Deng (邓迅)
 |
|
About Me
I am a Ph.D. student at University of Science and Technology of China (USTC), where I am fortunate to be advised by Professor Xiangnan He and Fuli Feng. Before that, I obtained a bachelor's degree in Electronic Information Engineering from USTC.
My research primarily focuses on the application of active learning across various fundamental machine learning tasks, aiming to explore efficient methods for selecting high-quality samples and achieving stable and sample-efficient model training. Specifically, my research interests encompass supervised learning, unsupervised clustering, virtual screening, and post-training of large language models. In these areas, I particularly emphasize the efficient generalization capabilities of models within the active learning framework and am dedicated to enhancing model performance in complex scenarios. If you share common interests and would like to explore collaboration or simply have a discussion, feel free to contact me.
Publications
Test-time Adaptation: We study how to achieve more reliable and accurate predictions in the evaluation phase in the presence of distribution shifts.
Learning to Double-Check Model Prediction From a Causal Perspective
Xun Deng, Fuli Feng, Xiang Wang, Xiangnan He, Hanwang Zhang, Tat-Seng Chua
IEEE Transactions on Neural Networks and Learning Systems (TNNLS) 2023
Improving the Hit Rates of Virtual Screening by Active Learning from Bioactivity Feedback
Xun Deng, Junlong Liu, Zhike Liu, Jiansheng Wu, Jieping Ye, Fuli Feng, and Zheng Wang
Major Revision at ACS Journal of Chemical Theory and Computation
Active Learning for Efficient Model Training: We study how to select high-quality samples for annotation under Out-of-Distribution (OOD) scenarios.
Counterfactual active learning for out-of-distribution generalization
Xun Deng, Wenjie Wang, Fuli Feng, Hanwang Zhang, Xiangnan He, Yong Liao
Association for Computational Linguistics (ACL) 2023
A3S: A General Active Clustering Method with Pairwise Constraints
Xun Deng, Junlong Liu, Han Zhong, Fuli Feng, Chen Shen, Xiangnan He, Jieping Ye, Zheng Wang
International Conference on Machine Learning (ICML) 2024
Less is More: Improving LLM Alignment via Preference Data Selection
Xun Deng, Han Zhong, Rui Ai, Fuli Feng, Zheng Wang, Xiangnan He
Others Related to Reinforcement Learning.
Risk-Sensitive Deep RL: Variance-Constrained Actor-Critic Provably Finds Globally Optimal Policy
Han Zhong, Xun Deng, Ethan X. Fang, Zhuoran Yang, Zhaoran Wang, Runze Li
Major Revision at Journal of the American Statistical Association (JASA)