Yuxuan Lu

Yuxuan Lu

Fourth-year Ph.D. student at Peking University working on Game Theory for AI, Information Elicitation, and Calibration for LLMs.

About Me

I am a fourth-year Ph.D. student at Peking University, advised by Prof. Yuqing Kong. My research focuses on applying game-theoretic principles to AI, especially large language models, with tools from information elicitation, mechanism design, and calibration theory.

I aim to bridge rigorous theory and practical deployment by developing mechanisms that are both provably sound and effective in real-world AI systems.

News

I served as a workshop organizer at WINE 2024, where Shengwei Xu and I presented the tutorial "Information Elicitation Meets Large Language Models: Progress, Opportunities, and Challenge." Materials are available here.

Education

Peking University, School of Computer Science

  • Sep 2022 - Jun 2027 (expected)
  • Ph.D. candidate, advised by Prof. Yuqing Kong
  • Research directions: Game Theory, Information Elicitation, Large Language Models

Peking University, School of EECS

  • Sep 2018 - Jun 2022
  • B.S. in Computer Science, Summa Cum Laude, Member of Turing Class

Selected publications

Game Theory for AI

  1. Jailbreaking LLMs via Calibration arXiv:2602.00619
    Authors: Yuxuan Lu, Yongkang Guo, Yuqing Kong.
    TL;DR: Reframes LLM jailbreaking as recalibration and proposes a gradient-shift attack to bypass safety guardrails.
    Materials: ArXiv
  2. Making and Evaluating Calibrated Forecasts arXiv:2510.06388
    Authors: Yuxuan Lu, Yifan Wu, Jason Hartline, Lunjia Hu.
    TL;DR: Designs truthful calibration metrics for evaluating AI model forecasts in a principled way.
    Materials: ArXiv
  3. Aligned Textual Scoring Rules arXiv:2507.06221
    Authors: Yuxuan Lu, Yifan Wu, Jason Hartline, Michael J. Curry.
    TL;DR: Develops LLM-embedded proper scoring rules to incentivize truthful textual judgments.
    Materials: ArXiv | PDF
  4. Benchmarking LLMs' Judgments with No Gold Standard ICLR 2025
    Authors: Shengwei Xu*, Yuxuan Lu*, Grant Schoenebeck, Yuqing Kong.
    TL;DR: Proposes a mutual-information-based metric to benchmark LLM judgments without ground truth.
    Materials: ArXiv | Proceedings
  5. Eliciting Informative Text Evaluations with Large Language Models EC 2024
    Authors: Yuxuan Lu*, Shengwei Xu*, Yichi Zhang, Yuqing Kong, Grant Schoenebeck.
    TL;DR: Designs a peer-prediction mechanism to incentivize informative free-form text evaluations from LLMs.
    Materials: ArXiv | PDF
  6. Calibrating "Cheap Signals" in Peer Review without a Prior NeurIPS 2023
    Authors: Yuxuan Lu, Yuqing Kong.
    TL;DR: Proposes a one-shot mechanism to calibrate biased peer review without prior data.
    Materials: ArXiv | PDF | Video | Slide

Side Topics (Blockchain, Entertainment)

  1. How Gold to Make the Golden Snitch: Designing the "Game Changer" in Esports arXiv:2405.19843
    Authors: Zhihuan Huang*, Yuxuan Lu*, Yongkang Guo, Yuqing Kong.
    TL;DR: Theorizes and empirically analyzes how to set rewards for a "game changer" item in esports.
    Materials: ArXiv
  2. A Framework of Transaction Packaging in High-throughput Blockchains arXiv:2301.10944
    Authors: Yuxuan Lu, Qian Qi, Xi Chen.
    TL;DR: Develops a game-theoretic framework for transaction packaging in high-throughput blockchains.
    Materials: ArXiv
  3. Empirical Analysis of EIP-1559: Transaction Fees, Waiting Time, and Consensus Security CCS 2022
    Authors: Yulin Liu, Yuxuan Lu, Kartik Nayak, Fan Zhang, Luyao Zhang, Yinhong Zhao.
    TL;DR: Conducts an empirical evaluation of Ethereum EIP-1559 on fees, waiting time, and consensus security.
    Materials: ArXiv | PDF | Proceedings
  4. FileInsurer: A Scalable and Reliable Protocol for Decentralized File Storage in Blockchain ICDCS 2022
    Authors: Hongyin Chen*, Yuxuan Lu*, Yukun Cheng.
    TL;DR: Proposes a decentralized blockchain file-storage protocol to improve both scalability and reliability.
    Materials: ArXiv | Proceedings
  5. SURPRISE! and When to Schedule It IJCAI 2021
    Authors: Zhihuan Huang, Shengwei Xu, You Shan, Yuxuan Lu, Yuqing Kong, Tracy Xiao Liu, Grant Schoenebeck.
    TL;DR: Quantifies how the timing of surprise in live esports affects audience-perceived quality.
    Materials: ArXiv | PDF | Proceedings

Profile

Experience

  • Visiting Predoctoral Fellow, Northwestern University (Feb 2025 - Jun 2025)
    Hosted by Prof. Jason Hartline
  • Research Intern, Duke University (Jun 2021 - Dec 2021)
    Hosted by Prof. Fan Zhang

Teaching

  • Teaching Assistant, Peking University
    Mathematical Foundations for the Information Age (Fall 2021, Fall 2022)
    Algorithm Design and Analysis (Spring 2021)

Academic Service

  • Program Committee Member: EC 2026
  • Workshop Organizer: WINE 2024
  • Reviewer: NeurIPS 2025, ICLR 2026, ICML 2026

Awards and Honors

  • Sep 2024: BYD Scholarship of Peking University
  • Sep 2021: John Hopcroft Scholarship of Peking University
  • Dec 2020: Gold Medal in 2020 ACM-ICPC Asia Regional Contest Shanghai Site
  • Sep 2020: Second-class Scholarship of Peking University
  • Dec 2019: Gold Medal in 2019 ACM-ICPC Asia Regional Contest Shanghai Site
  • Dec 2019: Gold Medal in 2019 ACM-ICPC Asia Regional Contest Shenyang Site
  • Sep 2019: Third-class Scholarship of Peking University
  • Dec 2018: Gold Medal in 2018 ACM-ICPC Asia Regional Contest Shenyang Site
  • Sep 2018: Freshman Scholarship of Peking University
  • Jul 2017: Gold Medal in the 34th National Olympiad in Informatics
  • Jul 2016: Gold Medal in the 33rd National Olympiad in Informatics