Howdy! Iām Weijia Zhang
I am an incoming M.S. student in Computer Science at Yale University (2026 - 2028), admitted to the Two-Year MS Degree with Full Scholarship.
I graduated from UIUC in Math + Computer Science with C.W. Gear Outstanding Undergraduate Student award (2 people per year).
Currently, I worked as a research assistant in UIUC U Lab on LLM agents, multimodal agents, and agentic RL, advised by Prof. Jiaxuan You.
News
- 2026.05: šš Graduated from UIUC in Math + Computer Science with the C.W. Gear Outstanding Undergraduate Student award! š
- 2026.01: šš Paper on bias inheritance was accepted to ACL 2026 as an Oral! šāØ
- 2025.10: š SeeingEye was released as an arXiv preprint and is under review at EMNLP 2026! šš
- 2025.07: š Joined Microsoft Research Asia as a research intern! š¬š¼
Research
My research interests center on LLM agents, especially next-generation AI agents that bridge virtual and physical worlds through socially intelligent, tool-agnostic, and ethically grounded architectures.
- Multimodal agents: memory, reasoning, tool use, and multi-agent systems
- Conversational AI: anthropomorphism and social intelligence
- Post-training: agent SFT and RL
Gamedev
Beyond research, I am a passoinate indie game developer, feel free to check my game work on the game page. I am also willing to discuss the future of AI X Game.
Publications
Research Articles
How Much Vision Does Multimodal Reasoning Need? Vision-Stripping for Multimodal Benchmarks
Published in Under review at NeurIPS 2026, 2026
Vision-stripping for multimodal benchmarks and multimodal reasoning.
Recommended citation: Weijia Zhang, Zijia Liu, Tianyi Zhang, Ruiqi Chen, Lian Zhang, Haoru Li, Haoqi Chen, and Jiaxuan You. (2026). "How Much Vision Does Multimodal Reasoning Need? Vision-Stripping for Multimodal Benchmarks." Under review at NeurIPS 2026.
Download Paper
Every Act Has Its Price: Compressed Moral Composition in Frontier LLMs
Published in Under review at EMNLP 2026, 2026
A study of compressed moral composition in frontier LLMs.
Recommended citation: Weijia Zhang, Ruiqi Chen, Yunze Xiao, and Weihao Xuan. (2026). "Every Act Has Its Price: Compressed Moral Composition in Frontier LLMs." Under review at EMNLP 2026.
Download Paper
CUADebug: Diagnosing and Repairing Computer-Use Agent Failures
Published in Under review at EMNLP 2026, 2026
A framework for diagnosing and repairing computer-use agent failures with a CUA-specific error taxonomy, benchmark, and tool-augmented debugger.
Recommended citation: Weijia Zhang et al. (2026). "CUADebug: Diagnosing and Repairing Computer-Use Agent Failures." Under review at EMNLP 2026.
Download Paper
SeeingEye: Agentic Information Flow Unlocks Multimodal Reasoning in Text-only LLMs
Published in arXiv preprint arXiv:2510.25092; under review at EMNLP 2026, 2025
Agentic information flow for unlocking multimodal reasoning in text-only LLMs.
Recommended citation: Weijia Zhang*, Zijia Liu*, Haoru Li*, Haoqi Chen*, and Jiaxuan You. (2025). "SeeingEye: Agentic Information Flow Unlocks Multimodal Reasoning in Text-only LLMs." arXiv preprint arXiv:2510.25092. Under review at EMNLP 2026.
Download Paper
Where LLM Agents Fail and How They Can Learn From Failures
Published in arXiv preprint arXiv:2509.25370, 2025
A study of LLM agent failures and how agents can learn from failed trajectories.
Recommended citation: Kunlun Zhu, Zijia Liu, Bingxuan Li, Muxin Tian, Yingxuan Yang, Jiaxun Zhang, Pengrui Han, Qipeng Xie, Fuyang Cui, Weijia Zhang, Xiaoteng Ma, Xiaodong Yu, Gowtham Ramesh, Jialian Wu, Zicheng Liu, Pan Lu, James Zou, and Jiaxuan You. (2025). "Where LLM Agents Fail and How They Can Learn From Failures." arXiv preprint arXiv:2509.25370.
Download Paper
Conference Papers
Understanding and Mitigating the Bias Inheritance in LLM-based Data Augmentation on Downstream Tasks
Published in ACL 2026, Oral, 2026
A systematic study of how bias can be inherited through LLM-generated synthetic data and how mitigation strategies behave across tasks.
Recommended citation: Miaomiao Li, Hao Chen, Yang Wang, Tingyuan Zhu, Weijia Zhang, Kaijie Zhu, Kam-Fai Wong, and Jindong Wang. (2026). "Understanding and Mitigating the Bias Inheritance in LLM-based Data Augmentation on Downstream Tasks." ACL 2026. Oral.
Download Paper
Artificial Leviathan: Exploring Social Evolution of LLM Agents Through the Lens of Hobbesian Social Contract Theory
Published in The First Workshop on AI Behavioral Science, ACM SIGKDD 2024, 2024
A simulated LLM agent society for studying emergent social contracts through Hobbesian social contract theory.
Recommended citation: Gordon Dai*, Weijia Zhang*, Jinhan Li, Siqi Yang, Srihas Rao, Arthur Caetano, and Misha Sra. (2024). "Artificial Leviathan: Exploring Social Evolution of LLM Agents Through the Lens of Hobbesian Social Contract Theory." The First Workshop on AI Behavioral Science, ACM SIGKDD 2024.
Download Paper
Education
Yale University
M.S. in Computer Science
2026-08 - 2028-05
Thesis Track with Full Scholarship
University of Illinois Urbana-Champaign
B.S. in Computer Science and Mathematics
2022-08 - 2026-05
GPA: 3.7/4.0
Work Experience
Microsoft Research Asia, Microsoft
Research Intern
Jul 2025 - Sep 2025
Worked on VLM/LLM agent research to improve Microsoft Excel Copilot capabilities.
- Built the TextAnalysisSFT data pipeline for SFT data generation for the new TextAnalysis API in Excel Copilot.
- Mined 2000+ real Kaggle samples, filtered heavy-text sheets, generated queries and Office.js code, and validated outputs with SheetEngine.
- Delivered a dataset that improved Office Script code-generation accuracy by 75%.
Reborn Network
AI Engineer
May 2023 - Jul 2023
Built real-time role-playing agent systems in Unity VR.
- Developed a role-playing agent Unity VR game enabling agents to interact through text, voice, and VR actions in real time with under 1s latency.
- Introduced RAG and vector databases to strengthen long-term agent memory, improving dialogue coherence score from 2/5 to 4/5.
- Designed a reusable character-card framework, enabling a UGC ecosystem and reducing character persona configuration time by 300%.
Tencent, WeChat Group
Software Engineer
Aug 2024 - Sep 2024
Built performance analysis tooling for WeChat Mini Programs and Unity memory profiling.
- Developed a cross-platform Android and iOS hardware performance analysis tool for WeChat Mini Programs, supporting 200+ partner teams in identifying performance bottlenecks.
- Built a Unity Mono Memory Profiler that discovered 40+ hidden memory allocation points, reducing memory-leak-related crash rate by 120%.