About

I am a Research Scientist at Google DeepMind. My research interests lie in Machine Learning, including

• Reinforcement learning from human feedback (RLHF), offline preference-based reinforcement learning
• Trustworthy Reinforcement Learning: Adversarial Attack, Provably Efficient Exploration, Provably Robust Exploration, and Verification on Deep Reinforcement Learning (DRL)
• Multi-Arm Bandit Learning Theories (MAB)

Education

University of Illinois Urbana-Champaign

• PhD in Computer Science, 2021 - 2026
• Advisor: Prof. Singh Gagandeep

Georgia Institute of Technology

• M.S. in Electrical and Computer Engineering, 2019 - 2021
• Advisor: Prof. Jacob Abernethy

Peking University

• B.S. in Physics, 2015 - 2019
• Advisor: Prof. Yun-Feng Xiao

Work Experience

• Google DeepMind, Research Scientist, 2025 - present
• Meta, Machine Learning Engineer Intern, 2025 Summer
• Google, Research Intern, 2024 Summer
• Amazon, Applied Scientist Intern, 2023 Summer
• Amazon, Applied Scientist Intern, 2022 Summer

Publications

[Google Scholar]. * means equal contribution.

2026

• Latent Adversarial Regularization for Offline Preference Optimization
  Enyi Jiang, Yibo Jacky Zhang, Yinglun Xu, Andreas Haupt, Nancy Amato, Sanmi Koyejo
  arXiv:2601.22083
  [PDF]

• Robust Thompson Sampling Algorithms Against Reward Poisoning Attacks
  Yinglun Xu, Zhiwei Wang , Gagandeep Singh
  AISTATS 2026
  [PDF]

2025

• Learning a Pessimistic Reward Model in RLHF
  Yinglun Xu, Hangoo Kang, Tarun Suresh, Yuxuan Wan, Gagandeep Singh
  arXiv: 2505.20556
  [PDF]

• Improving Assembly Code Performance with Large Language Models via Reinforcement Learning
  Anjiang Wei, Tarun Suresh, Huanmi Tan, Yinglun Xu, Gagandeep Singh, Ke Wang, Alex Aiken
  arXiv: 2505.11480
  [PDF]

2024

• Binary Reward Labeling: Bridging Offline Preference and Reward-Based Reinforcement Learning
  Yinglun Xu*, David Zhu*, Rohan Gumaste, Gagandeep Singh
  arXiv:2406.10445
  [PDF]

• Universal Black-Box Reward Poisoning Attack against Offline Reinforcement Learning
  Yinglun Xu*, Rohan Gumaste*, Gagandeep Singh
  arXiv:2402.09695
  [PDF]

• Two-Step Offline Preference-Based Reinforcement Learning with Constrained Actions
  Yinglun Xu, Tarun Suresh, Rohan Gumaste, David Zhu, Ruirui Li, Zhengyang Wang, Haoming Jiang, Xianfeng Tang, Qingyu Yin, Monica Xiao Cheng, Qi Zeng, Chao Zhang, Gagandeep Singh
  arXiv:2401.00330
  [PDF]

2023

• Efficient Reward Poisoning Attacks on Online Deep Reinforcement Learning
  Yinglun Xu, Qi Zeng, Gagandeep Singh
  Transactions on Machine Learning Research (TMLR)
  [PDF] [Code]

• On the robustness of epsilon greedy in multi-agent contextual bandit mechanism
  Yinglun Xu, Bhuvesh Kumar, Jacob Abernethy
  arXiv:2307.07675
  [PDF]

• Black-Box Targeted Reward Poisoning Attack Against Online Deep Reinforcement Learning
  Yinglun Xu, Gagandeep Singh
  arXiv:2305.10681
  [PDF]

2022

• Single-molecule optofluidic microsensor with interface whispering gallery modes
  Xiao-Chong Yu, Shui-Jing Tang, Wenjing Liu, Yinglun Xu, Qihuang Gong, You-Ling Chen, Yun-Feng Xiao
  PNAS 2022 (Proceedings of the National Academy of Sciences)
  [PDF]

• Thin-film optical parametric oscillators
  Alireza Marandi, Luis Ledezma, Yinglun Xu, Ryan Briggs
  US Patent 2022
  [PDF]

2021

• Observation-Free Attacks on Stochastic Bandits
  Yinglun Xu, Bhuvesh Kumar, Jacob Abernethy
  NeurIPS 2021
  [PDF]

• Adversarial Attack and Robust Learning in Multi-Arm Bandit Problems
  Yinglun Xu
  M.S. Thesis
  [PDF]

2020

• Bridging Truthfulness and Corruption-Robustness in Multi-Armed Bandit Mechanisms
  Jacob Abernethy, Bhuvesh Kumar, Thodoris Lykouris, Yinglun Xu (Alphabetically ordered)
  ICML 2020 Incentives in Machine Learning Workshop
  [PDF]

2018

• Mode splitting induced by an arbitrarily shaped Rayleigh scatterer in a whispering-gallery microcavity
  Yinglun Xu, Shui-jing Tang, Xiaochong Yu, Yi-Lin Chen, Daquan Yang, Qihuang Gong, Yun-Feng Xiao
  PRA 2018 (Physical Review A)
  [PDF]