News
26 Sep 2024
One full paper is accepted by NeurIPS'24 on alignment.
20 Sep 2024
One full paper is accepted by EMNLP'24 on agent.
02 Feb 2024
One full paper is accepted by WWW'24 on recommendation.
30 Nov 2023
Two full papers are accepted by ICDE'24 on recommendation and Graph.
22 Sep 2023
One full paper is accepted by NeurIPS'23 on Contrastive Learning.
20 April 2023
One full paper is accepted by IJCAI'23 on invariant learning for unbiased recsys.
26 Jan 2023
Two full papers are accepted by WWW'23 on recsys.
08 Aug 2021
One full paper is accepted by CIKM'21 on knowledge graph.
|
Junkang Wu
PhD student
Lab of Data Science 443 Huangshan Road, Hefei, China 230027
Advisor: Prof. Xiang Wang |
I am currently in the 3rd year of my PhD program at USTC Lab for Data Science, supervised by Prof. Xiang Wang and Prof. Xiangnan He.
My research interests includes llm alignment and robust optimization, particularly in direct preference optimization (DPO).
Education
|
University of Science and Technology of China (USTC) PhD Candidate in Cyberspace Security 2021.09 - 2026.06 (Expected) Advisor: Prof. Xiang Wang and Prof. Xiangnan He |
|
Zhejiang University of Technology (ZJUT) Bachelor in School of Computer Sciences 2017.09 - 2021.06 Advisor: Prof. Jianping Mei |
Preprints
 |
Towards Robust Alignment of Language Models: Distributionally Robustifying Direct Preference Optimization Junkang Wu, Yuexiang Xie, Zhengyi Yang, Jiancan Wu, Jiawei Chen, Jinyang Gao, Bolin Ding, Xiang Wang, Xiangnan He arxiv 2024 [PDF] [Codes] |
Publications
In the Year of 2024: |
$\beta$-DPO: Direct Preference Optimization with Dynamic $\beta$ Junkang Wu, Yuexiang Xie, Zhengyi Yang, Jiancan Wu, Jinyang Gao, Bolin Ding, Xiang Wang, Xiangnan He NeurIPS 2024 [PDF] [Codes] |
|
Direct Multi-Turn Preference Optimization for Language Agents Wentao Shi, Mengqi Yuan, Junkang Wu, Qifan Wang, Fuli Feng EMNLP 2024 [PDF] |
|
Lower-Left Partial AUC: An Effective and Efficient Optimization Metric for Recommendation Wentao Shi, Wang Chenxu, Fuli Feng, Yang Zhang, Wenjie Wang, Junkang Wu, Xiangnan He WWW 2024 (Full, Accept rate: 20.2%) [PDF] [Codes] |
|
BSL: Understanding and Improving Softmax Loss for Recommendation Junkang Wu, Jiawei Chen, Jiancan Wu, Wentao Shi, Jizhi Zhang, Xiang Wang ICDE 2024 (Full) [PDF] [Codes] |
|
Masked Graph Modeling with Multi-View Contrast Yanchen Luo, Sihang Li, Yongduo Sui, Junkang Wu, Jiancan Wu, Xiang Wang, Xiangnan He ICDE 2024 (Full) |
|
Understanding Contrastive Learning via Distributionally Robust Optimization Junkang Wu, Jiawei Chen, Jiancan Wu, Wentao Shi, Xiang Wang, Xiangnan He NeurIPS 2023 (Full, Accept rate: 26.1%) [PDF] [Codes] [Slides] [知乎] |
|
Adap-$\tau$: Adaptively Modulating Embedding Magnitude for Recommendation Jiawei Chen*, Junkang Wu*, Jiancan Wu, Sheng Zhou, Xuezhi Cao, Xiangnan He WWW 2023 (Full, Accept rate: 19.2%) [PDF] [Codes] [Slides] [知乎] * Co-first author |
|
On the Theories Behind Hard Negative Sampling for Recommendation Wentao Shi, Jiawei Chen, Fuli Feng, Jizhi Zhang, Junkang Wu, Chongming Gao, Xiangnan He WWW 2023 (Full, Accept rate: 19.2%) [PDF] [Codes] [Slides] |
|
Discriminative-Invariant Representation Learning for Unbiased Recommendation Hang Pan, Jiawei Chen, Fuli Feng, Wentao Shi, Junkang Wu, Xiangnan He IJCAI 2023 (Full, Accept Rate: 15.0%) [PDF] [Codes] |
|
DisenKGAT: Knowledge Graph Embedding with Disentangled Graph Attention Network Junkang Wu, Wentao Shi, Xuezhi Cao, Jiawei Chen, Wenqiang Lei, Fuzheng Zhang, Wei Wu, Xiangnan He CIKM 2021 (Full, Accept rate: 21.3%) [PDF] [Codes] [Slides] [知乎] |
Experiences
| Research Intern, Alibaba DAMO Academy, Hangzhou, Oct 2023 - Now Mentor: Yuexiang Xie |
| Research Intern, Meituan NLP Group, Beijing, Aug 2020 - Aug 2021 Mentor: Xuezhi Cao |
Honors
|
National Scholarship, 2021.12 - University of Science and Technology of China |
|
Outstanding graduates, 2021.06 - Zhejiang Province |
|
National Scholarship, 2019.12 - Zhejiang University of Technology |
Useful Links
| Machine Learning Reading List |
| Deep Learning Reading List |
Webpage template borrows from Xiangnan He.