Mofei Chen

Undergraduate student in Computer Science at Beijing Institute of Technology.

I am an undergraduate student in Computer Science at Beijing Institute of Technology, currently visiting the Knowledge Engineering Group (KEG) at Tsinghua University.

My current interests center on large language models and autonomous agents, especially automated evaluation for interactive UI systems, closed-loop agent workflows, diffusion-based generation systems, and reinforcement-learning-based decision agents.

I am actively preparing for graduate study and research opportunities. Outside this more formal homepage, I keep a personal blog at Timothy’s Blog.

Research Interests

  • Large language models and autonomous agents
  • Automated UI and interactive courseware evaluation
  • Multimodal generation and diffusion models
  • Reinforcement learning for decision-making agents

News

  • Jan. 2026 - Present: Visiting student at THU-KEG, working on agent-as-a-judge evaluation for interactive courseware.
  • 2026: Co-first author paper on ICW-Bench is under submission to EMNLP 2026.
  • Apr. 2026: Won Provincial Second Prize in the Lanqiao Cup Agent Development Track.
  • Nov. 2025: Won National First Prize, 3rd Place, in the DataCon Big Data Security Analytics Competition.
  • Apr. 2025: Won National First Prize, 1st Place, in the Westlake Cybersecurity Competition Security Agent Track.
  • 2024-2025: Received the National Scholarship and BIT First-Class Academic Scholarships.

Education

Beijing Institute of Technology, School of Computer Science
B.S. in Computer Science and Technology, expected 2027
GPA: 3.75/4.00; Average score: 89.73; Rank: 16/115

Experience

Tsinghua University, Knowledge Engineering Group (THU-KEG)
Visiting Student, Agent-as-a-judge and Browser Use
Jan. 2026 - Present

I work on automated evaluation for interactive courseware, focusing on browser-agent exploration, multimodal judge validation, and feedback annotation for ICW-Bench.

Beijing Institute of Technology, School of Computer Science
Project Lead, AIGC and Image Stylization
Nov. 2024 - Oct. 2025

I led the design and backend implementation of a multimodal image-generation platform that connects LLM-based intent parsing, workflow planning, and AIGC inference.

Selected Projects

ICW-Bench: Evaluating Interactive Courseware with Agent-as-a-judge

Co-first author work under submission to EMNLP 2026.

We propose an Interact-then-Review framework for evaluating interactive courseware, where a browser agent first explores the UI and records interaction trajectories, then a multimodal judge reviews key frames and dynamic evidence. I contributed to expert metric design, multimodal data collection and cleaning, and the browser-agent evaluation pipeline.

LLM-Driven Automated Rule Synthesis for Security Logs

National First Prize, 1st Place, Westlake Cybersecurity Competition Security Agent Track.

This project converts LLM semantic understanding into efficient offline parsing rules for heterogeneous security logs. The system uses semantic clustering, feedback-driven code generation, and boosting-inspired residual coverage to build high-performance parsing rules without runtime LLM dependency.

Multimodal Image Style Transfer Platform

BIT undergraduate innovation project.

I built an end-to-end platform that uses LLMs as controllers to parse text and image inputs, plan generation workflows, and execute backend AIGC inference. The project deepened my understanding of diffusion models, Flow Matching, asynchronous backend architecture, and multimodal product design.

Rule-Guided Reinforcement Learning for MOBA Decision Agents

Course project, ranked 2nd in class with a score of 97/100.

This project studies how rule guidance can improve PPO exploration in high-dimensional MOBA action spaces. I designed condition-triggered macro-action pruning and multi-objective rewards for positioning and lane-clearance behavior.

Honors & Awards

  • National Scholarship, top 5%.
  • Beijing Institute of Technology Outstanding Student Model, top 3%.
  • BIT First-Class Academic Scholarship, consecutive recipient.
  • Westlake Cybersecurity Competition, Security Agent Track, National First Prize, 1st Place, 2025.
  • DataCon Big Data Security Analytics Competition, National First Prize, 3rd Place, 2025.
  • Chinese Mathematics Competitions, National Second Prize, 2024.
  • Lanqiao Cup Agent Development Track, Provincial Second Prize, 2026.