Jiatao Li

Jiatao Li

Undergrad student at Peking University working on NLP

About me

I’m a junior student from Information Management Department, Peking University. I am fortunate to work as a research assistant at Wangxuan Institute of Computer Technology under the supervision of Prof. Xiaojun Wan. In summer 2025, I will work as an undergraduate visiting researcher (UGVR) at Stanford University. My research interest lies broadly in NLP and ML. Recently I am drawn by the development of large language models. A list of topics i am actively thinking about/working on:

  • Enhancing Knowledge Retrieval in LLMs: I am deeply engaged in exploring methodologies that can dynamically enhance large language models’ ability to retrieve information, akin to human cognition. I believe retrieval is crucial for enabling LLMs to continually learn and adapt to the ever-changing world knowledge.

  • Human-LLM Interaction Dynamics: Another facet of my research curiosity lies in examining the interfaces between large language models and their human users. I am particularly captivated by the concept of a highly tailored AI assistant capable of evolving in tandem with a user’s individual experiences and memories, thereby redefining personalized interaction.

Through these pursuits, I strive to contribute to the forefront of NLP and ML research, pushing the boundaries of what’s possible in artificial intelligence while drawing insights that bridge the gap between machine and human cognition. My long-term research goal is to build a self-evolving LLM capable of continuous learning and adaptation, ultimately achieving a higher degree of artificial general intelligence.

You can find my CV here: Jiatao Li’s Curriculum Vitae.

Preprints

  • SMART-RAG: Selection using Determinantal Matrices for Augmented Retrieval
    Jiatao Li, Xinyu Hu, Xiaojun Wan
    arXiv:2409.13992 [pdf]

  • Aspect-Guided Multi-Level Perturbation Analysis of Large Language Models in Automated Peer Review
    Jiatao Li, Yanheng Li, Xinyu Hu, Mingqi Gao, Xiaojun Wan
    arXiv:2502.12510 [pdf]

  • Who Writes What: Unveiling the Impact of Author Roles on AI-generated Text Detection
    Jiatao Li, Xiaojun Wan
    arXiv:2502.12611 [pdf]

Selected Publications

  • Evaluating Self-Generated Documents for Enhancing Retrieval-Augmented Generation with Large Language Models
    Jiatao Li, Xinyu Hu, Xunjian Yin, Xiaojun Wan
    NAACL 2025 (Findings). (Long Paper) [pdf]

Selected Awards

  • 2023-2024: Peking University Merit Student Pacesetter
  • 2024: Leo KoGuan Scholarship
  • 2024: CSDN Blog Star of the Year Top87
  • 2023: Peking University Annual Person of the Year Nomination (Youngest Nominee, 45/65k+, blog spotlight in Chinese)
  • 2023: Peking University May 4th Scholarship (Highest-level Scholarship in Peking University, 125/65k+)
  • 2023: CSDN Blog Star (Beijing) Top10
  • 2023: Kaggle Competition:HousePrice Prediction Top1%
  • 2023: Tianchi BMW 3rd Hackathon - Virtual Collaborative Co-Creation Track & Finals Rank1
  • 2023: Peking University Hackathon Track2(WebGPU) Rank2
  • 2023: Greater Bay Area (Huangpu) International Algorithm Case Competition - Advanced Enhancement of Large Language Model Comprehensive Abilities (Pazhou Laboratory & Fudan University), Finalist.