Wenqing Tian

Email: is_wenqingtian@outlook.com

Education

Beijing Institute of Technology | September 2022 – Present School of Computer Science & Technology

• GPA Ranking (First Five Semesters): No.1/117 (91.51/100)
• Overall Ranking (First Five Semesters): No.1/117
• CET-6: 561
• Relevant Courses: Programming Methodology and Practice (100), Operating Systems (99), Machine Learning Theory and Practice (99), Linear Algebra (99), Reinforcement Learning (98), Assembly Language and Interface Technology (97), Digital Logic (97), Object-Oriented Technology and Methods (97), Algorithmic Game Theory (97), Introduction to Natural Language Understanding (95), Fundamentals of Web Development (95), Special Topics in New Computer Technology (95), Data Structures and Algorithm Design (94), C Language Programming (94)

Research Experience

Efficient Exploration for R1-Searcher Follow-up

• Institution: Renmin University of China
• Date: March 2024 – Present
• Supervisor: Prof. Xin Zhao (AIBox RUC)
• Collaborators: Huatong Song, Jinhao Jiang
• Description:
  • Existing models trained via reinforcement learning to autonomously call external search engines show a tendency to over-rely on these engines.
  • We are dedicated to developing models that can efficiently call external search engines while appropriately leveraging their internal knowledge to generate high-quality responses.
  • Participated in cold-start phase exploration, reward function design, and model rewriting training.

Exploration of Attacks and Defenses for Retrieval-Augmented Generation (RAG) Systems

• Institution: Institute of Computing Technology, Chinese Academy of Sciences
• Date: December 2024 – February 2025
• Supervisor: Prof. Fei Sun (Personal Page)
• Description:
  • We are exploring mechanisms to efficiently defend against database poisoning attacks.
  • We are attempting to effectively utilize the internal knowledge of large language models in defense mechanisms.

Ancient Chinese Poetry Q&A System based on Retrieval-Augmented Generation (RAG)

• Institution: Beijing Institute of Technology
• Date: October 2024 – November 2024
• Supervisor: Prof. Yang Gao (Faculty Page)
• Description:
  • Responsible for implementing the Retrieval-Augmented Generation module, performing intent recognition for user queries, and dynamically invoking RAG:
    • For fact-based queries, invoke RAG to improve the accuracy of model responses;
    • For creative queries, invoke specific system prompts to maintain the creativity of model responses.

Optimization of Approximate Maximin Share (MMS) for Indivisible Goods in Fair Division

• Institution: Beijing Institute of Technology
• Date: September 2024 – November 2024
• Supervisor: Prof. Zhengyang Liu (Personal Page)
• Description:
  • We are simultaneously studying two approximation models: multiplicative approximation and ordinal approximation:
    • In multiplicative approximation, we aim to improve the lower bound of (\alpha) in (\alpha)-MMS by enhancing the bagging method;
    • In ordinal approximation, we aim to explore whether the upper bound of (d) in (1)-out-of-(d) MMS can be further constrained when all agent valuation functions are identical.

Interference Fringe Parameter Estimation Based on Huawei Cloud

• Institution: Beijing Institute of Technology (Project Lead)
• Date: December 2023 – November 2024
• Supervisor: Prof. Mingfeng Lu (Faculty Page)
• Description:
  • Utilized deep learning techniques to analyze Newton’s rings interference phenomena, achieving regression from a single interference fringe image to obtain parameters such as the coordinates of Newton’s ring center and the lens’s radius of curvature.
  • Improved the prediction accuracy of the two geometric parameters, ring center coordinates and radius of curvature, through measures such as task decoupling, feature sharing, independent optimization, and loss weight adjustment.
  • Implemented parameter estimation using PyTorch and MindSpore successively, and deployed on ModelArts, achieving separation of data acquisition and data processing.

Honors and Awards

• 2023 – 2024 Academic Year: First-Class Academic Scholarship, Xiaomi Special Scholarship, University Outstanding Student Pacesetter, Department Model of Academic Excellence and Integrity
• 2022 – 2023 Academic Year: First-Class Academic Scholarship, Xiaomi Scholarship, University Outstanding Student, University Outstanding League Member, University Outstanding Individual in Social Practice

Competition Experience

• National Collegiate Computer Artificial Intelligence Challenge (2024)
  • Award: Third Prize
  • Level: Provincial/Ministerial
• Mathematical Contest in Modeling (2024)
  • Award: Third Prize
  • Level: National
• The 5th Beijing Collegiate Physics Tournament (2024)
  • Award: Second Prize
  • Level: Provincial/Ministerial
• “FLTRP·ETIC Cup” English Comprehensive Ability Competition, Beijing Division (2024)
  • Award: Third Prize
  • Level: Provincial/Ministerial