How Much Vision Does Multimodal Reasoning Need? Vision-Stripping for Multimodal Benchmarks
Weijia Zhang, Zijia Liu, Tianyi Zhang, Ruiqi Chen, Lian Zhang, Haoru Li, Haoqi Chen, and Jiaxuan You
Under review
CV
Education
- M.S. in Computer Science, Yale University, Sep 2026 - May 2028
- Thesis Track with Full Scholarship
- Incoming student
- B.S. in Computer Science and Mathematics, University of Illinois Urbana-Champaign, Aug 2022 - May 2026
- GPA: 3.7/4.0
- 2026 C.W. Gear Outstanding Undergraduate Student, one of two annual recipients
- 2025 Dean’s List
Work Experience
- Research Scientist Intern, Multimodal LLMs/Agents, TikTok, Jun 2026 - Present
- Designed a self-evolving content-moderation agent framework centered on a ReAct agent that autonomously analyzes individual videos using specialized tools for frame sampling, ASR/OCR, metadata extraction, originality scoring, and watermark/celebrity detection; integrated multimodal evidence with policy-grounded reasoning to produce structured moderation decisions.
- Built a self-evolution loop where a Reflector aggregates Solver failures into root-cause attribution reports and an Evolver adds, revises, or merges case-relevant skills to refine the Solver, improving the overall F1 score by 15.5%.
- Research Scientist Intern, Microsoft, Jul 2025 - Sep 2025
- Conducted LLM research for Microsoft 365 Excel Copilot, focusing on spreadsheet text analysis and Office.js code generation.
- Developed the TextAnalysisSFT data pipeline for Excel Copilot’s TextAnalysis API: mined 2,000+ real-world examples from Kaggle, filtered text-heavy spreadsheets, generated user queries and Office.js programs, and validated executions with SheetEngine.
- The resulting dataset improved Office Scripts code-generation accuracy by 15%.
- AI Engineer, Reborn Network, May 2023 - Jul 2023
- 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%.
- Software Engineer, WeChat Group, Tencent, Aug 2024 - Sep 2024
- 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%.
Publications
Every Act Has Its Price: Compressed Moral Composition in Frontier LLMs
Weijia Zhang*, Ruiqi Chen*, Yunze Xiao*, and Weihao Xuan
EMNLP 2026
CUADebug: Diagnosing and Repairing Computer-Use Agent Failures
Weijia Zhang, Kunlun Zhu, Zeyi Liu, Yinting Chen, Tianyi Ma, Jiateng Liu, Jiaxun Zhang, Bingxuan Li, Pan Lu, Xiangru Tang, Heng Ji, and Jiaxuan You
EMNLP 2026
Understanding and Mitigating the Bias Inheritance in LLM-based Data Augmentation on Downstream Tasks
Miaomiao Li, Hao Chen, Yang Wang, Tingyuan Zhu, Weijia Zhang, Kaijie Zhu, Kam-Fai Wong, and Jindong Wang
ACL 2026, Oral
SeeingEye: Agentic Information Flow Unlocks Multimodal Reasoning in Text-only LLMs
Weijia Zhang*, Zijia Liu*, Haoru Li*, Haoqi Chen*, and Jiaxuan You
arXiv:2510.25092 · AAAI 2026 under review
Where LLM Agents Fail and How They Can Learn From Failures
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
ICML 2026 Workshop
Artificial Leviathan: Exploring Social Evolution of LLM Agents Through the Lens of Hobbesian Social Contract Theory
Gordon Dai*, Weijia Zhang*, Jinhan Li, Siqi Yang, Srihas Rao, Arthur Caetano, and Misha Sra
The First Workshop on AI Behavioral Science, ACM SIGKDD 2024
Project Experience
- OpenManus & OpenManus-RL, Researcher, Jan 2025 - Jul 2025
- Collaborated with the official OpenManus team as a core contributor to OpenManus-RL, the follow-up project of OpenManus.
- The open-source ecosystem has accumulated 60,000+ GitHub stars.
- Implemented a ReAct-style reasoning architecture and multi-agent collaboration workflows.
- Built an MCP-based tool-library framework for automated tool invocation and output processing.
- Constructed high-quality SFT datasets for agent benchmarks including AlfWorld and WebShop.
- Designed step-level reward signals to improve agent planning and memory capabilities.
- SeeingEye, Co-first Author / Researcher, Oct 2025 - Present
- Proposed an agentic information-flow framework that enables text-only LLMs to perform multimodal reasoning through structured, tool-mediated textual context.
- Designed the perception-to-reasoning pipeline for multimodal tasks while keeping the backbone model text-only.
- Studied how tool use, intermediate representations, and memory-like context unlock multimodal reasoning behavior in text-only LLMs.
- Released the work as an arXiv preprint, currently under review at EMNLP 2026.
- Paper: SeeingEye
- GUIAgentDebugger, First Author / Lead, Jan 2026 - Present
- Designed a GUI-agent error taxonomy with 4 major categories and 29 subtypes across perception, interaction localization, task reasoning, and external system failures.
- Built a self-evolving VLM-agent debugging framework that identifies root causes from failed trajectories, distills them into reusable debugging skills, and enables agents to learn from historical failures.
- Designed a dual-layer memory architecture with episodic and semantic memory, plus intent-aware RAG retrieval for similar-intent trajectory skills.
Activities and Awards
- May 2026: C.W. Gear Outstanding Undergraduate Student, one of two annual recipients
- 2025: UIUC Dean’s List
- Aug 2025: RecSys 2025 Reviewer
- Sep 2022: UIUC MechMania 28 AI Hackathon, 2nd Place
Skills
- Programming languages: Python, C/C++, C#, Java, JavaScript/TypeScript, HTML/CSS, SQL, Rust
- Frameworks and libraries: VERL, VLLM, LangGraph, LangChain, PyTorch, TensorFlow
- AI focus: SFT, reinforcement learning, post-training, RAG, agentic AI, machine learning, generative AI