Photo taken by my friend Jiao in New Mexico
Embedded Files
๐ฎ nanx [AT] usc [DOT] edu
Hi! I am a Research Scientist in Google DeepMind.
I aim to build trustworthy LLM systems.ย Towards this , I investigateย
inference-time algorithms to improve LLM factuality
safety alignment of LLMs in defense of jailbreaks, adversarial attacks, and etc.
the design of multimodal LLMs, focusing on pre-training, fine-tuning, and in-context learning.
I obtained my Ph.D. degree at Department of Computer Science of University of Southern California, advised by ย Prof. Muhao Chen and ย Prof. Xuezhe Ma. Before coming to USC, I received my Bachelor's (in 2015) and Master's degree (in 2019) of Computer Science from Shanghai Jiao Tong University, advised by Prof. Yanmin Zhu and Prof. Yanyan Shen. I also got a second Master's degree (in 2017) of Computer Science from Technical University of Berlin. I was fortunate to work with Prof. Yan Liu and Prof. Zhenhui Li.
News
News
[Jun 26th, 2025] I passed thesis defense today ๐๐
[Jan 22nd, 2025] Two papers, From Introspection to Best Practices and LLM The Genius Paradox, were accepted by NAACL 2025
[Jan 22nd, 2025] Our benchmark for multi-image understanding, MuirBench, was accepted by ICLR 2025
[Mar 13th, 2024] Our paper Cognitive Overload was accepted by NAACL 2024
[Jan 18th, 2024] Our paper CEO was accepted by EACL 2024
[Jan 16th, 2024] Our paper AutoDAN was accepted by ICLR 2024
Experience
Experience
Microsoft Research
May. 2024 - Aug. 2024 Redmond, U.S.A.
Research Intern, Mentors: Dr. Sheng Zhang and Dr. Hoifung Poon
In-context Learning for Multimodal Large Language Models
Tencent AI Lab
Jun. 2022 - Aug. 2022 Bellevue, U.S.A.
Research Intern, Mentors: Dr. Hongming Zhang and Dr. Jianshu Chen
Corpus-based Open-Domain Event Ontology Induction
Amazon
Jun. 2020 - Aug. 2020 Seattle, U.S.A.
Applied Science Intern, Mentors: Dr. Seyi Feyisetan and Dr. Abhinav Aggarwal
Differentially Private Textual Perturbations Using Truncated Gumbel Noise
Fraunhofer Heinrich Hertz Institute
Jun. 2016 - Sep. 2017 Berlin, Germany
Software Engineer Intern
Measuring Station Software Development and Maintenance (in C++)
Publications
Publications
(* indicates equal contribution)
Multimodal LLMs
Multimodal LLMs
Does (feature) alignment enable LLMs multimodal in-context learning?ย
Nan Xu, Tim Ossowski, Sheng Zhang, Muhao Chen, Hoifung Poon
In submission
MUIRBENCH: A Comprehensive Benchmark for Robust Multi-image Understanding ๐๐ก ๐งโ๐ป
Fei Wang, Xingyu Fu, James Y. Huang, Zekun Li, Qin Liu, Xiaogeng Liu, Mingyu Derek Ma, Nan Xu, Wenxuan Zhou, Kai Zhang, Tianyi Yan, Wenjie Jacky Mo, Hsiang-Hui Liu, Pan Lu, Chunyuan Li, Chaowei Xiao, Kai-Wei Chang, Dan Roth, Sheng Zhang, Hoifung Poon, Muhao Chen
ICLR 2025
From Introspection to Best Practices: Systematic Analysis of Demonstrations in Multimodal In-Context Learning ๐๐งโ๐ป
Nan Xu, Feiwang, Sheng Zhang, Hoifung Poon, Muhao Chen
NAACL 2025
mDPO: Conditional Preference Optimization for Multimodal Large Language Models ๐๐ก ๐งโ๐ป
Fei Wang, Wenxuan Zhou, James Y. Huang, Nan Xu, Sheng Zhang, Hoifung Poon, Muhao Chen
EMNLP 2024
Natural Language Generation
Natural Language Generation
LLM The Genius Paradox: A Linguistic and Math Expertโs Struggle with Simple Word-based Counting Problems ๐๐งโ๐ป๐น
Nan Xu, Xuezhe Ma
NAACL 2025
Monotonic Paraphrasing Improves Generalization of Language Model Prompting ๐๐งโ๐ป
Qin Liu, Fei Wang, Nan Xu, Tianyi Yan, Tao Meng, Muhao Chen
EMNLP 2024 Findings
DecoPrompt : Decoding Prompts Reduces Hallucinations when Large Language Models Meet False Premises ๐
Nan Xu, Xuezhe Ma
In submission
๐ Outstanding Paper Award
Look-back Decoding for Open-Ended Text Generation ๐ ๐นย
Nan Xu, Chunting Zhou, Asli Celikyilmaz, Xuezhe Ma
EMNLP 2023
Evaluate Large Language Models on Controlled Generation Tasks ๐
Jiao Sun*, Yufei Tian*, Wangchunshu Zhou*, Nan Xu*, Qian Hu, John Frederick Wieting, Xuezhe Ma
EMNLP 2023
Safety and Robustness in NLP
Safety and Robustness in NLP
Cognitive Overload: Jailbreaking Large Language Models with Overloaded Logical Thinking ๐ ๐งโ๐ป
Nan Xu, Fei Wang, Ben Zhou, Bang Zheng Li, Chaowei Xiao, Muhao Chen
NAACL 2024 Findings
Autodan: Generating stealthy jailbreak prompts on aligned large language models ๐๐งโ๐ป
Xiaogeng Liu, Nan Xu, Muhao Chen, Chaowei Xiao
ICLR 2024
๐ Best Paper Award
Density-Aware Differentially Private Textual Perturbations Using Truncated Gumbel Noise ๐
Nan Xu, Oluwaseyi Feyisetan, Abhinav Aggarwal, Zekun Xu, Nathanael Teissier
FLAIRS 2021
NLP Applications
NLP Applications
CEO: Corpus-based Open-Domain Event Ontology Induction ๐ ๐ก ๐งโ๐ป
Nan Xu, Hongming Zhang, Jianshu Chen
EACL 2024 Findings
Dense Retrieval as Indirect Supervision for Large-space Decision Making ๐ ๐งโ๐ป
Nan Xu, Fei Wang, Mingtao Dong, Muhao Chen
EMNLP 2023 Findings
Does Your Model Classify Entities Reasonably? Diagnosing and Mitigating Spurious Correlations in Entity Typing ๐ย
Nan Xu,ย Fei Wang, Bangzheng Li, Mingtao Dong, Muhao Chen
EMNLP 2022
Reinforcement Learning Applications
Reinforcement Learning Applications
Treatment Recommendation with Preference-based Reinforcement Learning ๐๐กย
Nan Xu, Nitin Kamra, Yan Liu
ICBK 2021
Toward A Thousand Lights: Decentralized Deep Reinforcement Learning for Large-Scale Traffic Signal Control ๐
Chacha Chen, Hua Wei, Nan Xu, Guanjie Zheng, Ming Yang, Yuanhao Xiong, Kai Xu, Zhenhui Li
AAAI 2020
MetaLight: Value-based Meta-reinforcement Learning for Online Universal Traffic Signal Control ๐
Xinshi Zang, Huaxiu Yao, Guanjie Zheng, Nan Xu, Kai Xu, Zhenhui Li
AAAI 2020
CoLight: Learning Network-level Cooperation for Traffic Signal Control ๐ ๐งโ๐ปย
Hua Wei*, Nan Xu*, Huichu Zhang, Guanjie Zheng, Xinshi Zang, Chacha Chen, Weinan Zhang, Yanmin Zhu, Kai Xu, Zhenhui Liย
CIKM 2019ย
Targeted Knowledge Transfer for Learning Traffic Signal Plans ๐
Nan Xu, Guanjie Zheng, Kai Xu, Yanmin Zhu, Zhenhui Li
PAKDD 2019
Diagnosing Reinforcement Learning for Traffic Signal Control ๐
Guanjie Zheng, Xinshi Zang, Nan Xu, Hua Wei, Zhengyao Yu, Vikash Gayah, Zhenhui Li
Arxiv
Others
Others
A multi-scale time-series dataset with benchmark for machine learning in decarbonized energy grids ๐ ๐งโ๐ป
Xiangtian Zheng*, Nan Xu*,ย Loc Trinh*, Dongqi Wu*, Tong Huang, S Sivaranjani, Yan Liu, Le Xie
Scientific Data 2022
Interpretability and fairness evaluation of deep learning models on MIMIC-IV dataset ๐ย
Chuizheng Meng*, Loc Trinh*, Nan Xu*,ย Yan Liuย
Scientific Reports 2022
Simulating Continuous-time Human Mobility Trajectories ๐ ๐ผ๏ธ
Nan Xu*, Loc Trinh*, Sirisha Rambhatla, Samuel Assefa, Jiahao Chen, Zhen Zeng, Yan Liuย
SimDL@ICLR 2021
A Multi-task Learning Framework for Automatic Early Detection of Alzheimer's ๐ ๐ผ๏ธ
Nan Xu, Yanyan Shen, Yanmin Zhu
DASFAA 2019 (short)
Attention-based Hierarchical Recurrent Neural Network for Phenotype Classification ๐
Nan Xu, Yanyan Shen, Yanmin Zhu
PAKDD 2019ย
Awards
Awards
Outstanding Paper Award, EMNLP 2023
Sentosa, Singapore 2023
Best Paper Award, the 34th International FLAIRS Conference
Miami, U.S.A 2021
Annenberg Graduate Fellowship, University of Southern California
Los Angeles, U.S.A 2019-2023
Scholarship of the German Academic Exchange Service, Technical University of Berlin
Berlin, Germany 2015-2016
Academic Excellence Scholarship, Shanghai Jiao Tong University
Shanghai, China 2013-2014
Services
Services
PC Member
AAAI (2023)
Conference Reviewer ย ย ย ย ย ย ย
ACL (2023), EMNLP (2023, 2022), EACL (2022), ICML (2022),ย
NeurIPS (2023-2021), ICLR (2024, 2022),ย
AMIA (2020)
Journal Reviewer ย ย ย ย ย ย
TPAMI (2021)
Miscellaneous
Miscellaneous
Nikki Xu is my kitty. His Chinese name is ๅพๅคงๅฎ. He is a handsome silver shaded british shorthair. Please check more photos of Nikki here ๐ฅฐ
Google Sites
Report abuse