About

I am a Ph.D. candidate at the Gaoling School of Artificial Intelligence, Renmin University of China, advised by Prof. Jun Xu. I study trustworthy large language models and agents, with a focus on mechanistic interpretability. I expect to graduate in June 2027 and am seeking research positions.

My research draws on cognitive neuroscience to map how capabilities are organized inside models, identify the mechanisms behind failures, and intervene during long-running tasks. My first-author work includes 13 papers at CCF A/B venues (10 at CCF A venues). I also hold four granted invention patents. Research recognition includes an ICLR Spotlight and Best Paper Awards at SIGIR-AP and CCIR.

Google Scholar Updated
Citations
2139
h-index
19
i10-index
22

Research & Plans

Full Research Program

My current program, Cognitive Neuroscience of AI Agents, connects behavioral experiments, representation analysis, and causal intervention. The work follows three questions:

How are cognitive capabilities organized?
NeuroCogMap connects sparse model features to functional parcels, capability hierarchies, and the localization of model failures. NeuroCogMap
What causes failures in reasoning, retrieval, and memory?
ReDeEP studies the balance between retrieved evidence and parametric knowledge. RHD examines reasoning trajectories; FPPS studies factual drift caused by personalized memory. Failure Mechanisms
How can an agent correct its behavior during a task?
I study metacognitive control for long-horizon agents, connecting process evaluation and hierarchical self-monitoring with evidence-grounded human oversight. Agents assess reasoning and evidence use to identify drift and guide reflection and correction. Human-in-the-loop monitoring makes consequential decisions, their evidence, and their implications available for user verification, allowing human judgment and feedback to guide subsequent execution. EBG Paper Metacognitive Control

Selected Publications

Full Publication List

Background

Education

  • Renmin University of China2022.09 - 2027.06 (expected)

    Ph.D. in Artificial Intelligence

    Gaoling School of Artificial Intelligence. Advisor: Prof. Jun Xu.

  • Nanyang Technological University2025.09 - 2026.03

    Visiting Student

    College of Computing and Data Science. Host: Prof. Yang Liu. Research on AI safety and model interpretability.

  • Beijing Jiaotong University2018.09 - 2022.06

    B.Eng. in Computer Science and Technology

    Ranked 1st out of 250 students.

Research Experience

  • Westlake University-Renmin University of China Joint Institute for Future Humanity2025.10 - present

    Young Researcher

  • Kuaishou2022.02 - 2025.06

    Research Intern

    Model Applications, Search Technology, and Foundation Models and Applications teams. Mentors: Xiaoxue Zang, Kai Zheng, and Yang Song.

Projects & Applications

  • Participated in national key R&D and general research projects, and collaborations with Kuaishou, Lenovo, ByteDance Seed, and Huawei.
  • Research and engineering outcomes have been applied in Kuaishou search and recommendation, legal case recommendation, ByteDance Seed AI search, and Lenovo e-commerce agents.
  • Founded Huishen Technology to develop a trustworthy mock-court prototype using legal reasoning, retrieval, evidence verification, and multi-agent collaboration. The project won First Prize in the 2025 national university AI and smart-governance innovation competition (135 teams), with demonstrations in judicial and educational settings.
Four Granted Invention Patents
  • Explainable legal case matching using inverse optimal transport ZL202210448566.5
  • Causal representation learning for legal case matching ZL202310463448.6
  • Resource recommendation and prediction-model training ZL202310781584.X
  • Non-IID data simulation using large-model agents ZL202510095786.8

Honors & Awards

  • 2024, 2026Graduate National Scholarship. Selected in 2025 as one of 100 national recipient representatives featured in People's Daily. News
  • 2025China Association for Science and Technology Youth Talent Support Program for Ph.D. Students.
  • 2026CCIR Best Paper Award.
  • 2024SIGIR-AP Best Paper Award.
  • 2026Linghang Dean's Scholarship, Renmin University of China.
  • 2022Beijing Outstanding Graduate.
  • 2022, 2023, 2025University Merit Student.
  • Renmin University of China Innovative Talent Development Program.

Teaching & Mentoring

Teaching assistant at Renmin University of China:

  • Introduction to Big Data AnalyticsFall 2022, 2023
  • Python ProgrammingSpring 2023, 2024
  • Intelligent Information RetrievalFall 2024

I have helped supervise 10 undergraduate and graduate students in problem formulation, experiments, and paper writing. One graduate published seven research papers and joined the Tencent Qingyun Program. Other outcomes include undergraduate first-author SIGIR publications and graduate students' first research projects.

I plan to teach programming, data analysis, and information retrieval, with modules on model interpretability and agents built around paper reading, replication, and open-ended experiments.

Academic Service

  • Reviewer for SIGIR, WWW, CIKM, ACL, NeurIPS, ICLR, and ICML; journal reviewing for TOIS and TASLP.
  • Operations lead for RUC Aseed+Lab and member of the academic operations team at the Institute for AI Governance.
  • Organizer for the Chinese AI and Law Challenge (CAIL); volunteer at SIGIR-AP 2024.

Talks

  • Accurate and Explainable Legal Case Matching. Renmin University of China's International Academic Symposium for Doctoral Students. Slides
  • Invited participant, Zhongguancun Science City Rixin Youth Talk Forum. News