Yuanda Wang

Yuanda Wang

I received my B.E. degree from Beijing University of Posts and Telecommunications(BUPT), Beijing, China, in 2024. I am currently working toward the master’s degree at Peking University. My research interest include system security and AI for security.

About Me

Master’s Student | Wang Xuan Institute of Computer Technology, Peking University (Sep 2024 - Present) My research focuses on AI for Security, Binary Analysis, and LLM Alignment and Reasoning Optimization. Deeply engaged in both academia and the tech industry, I am dedicated to leveraging artificial intelligence to solve fundamental system security challenges.

Academic & Research Background

Core Research Affiliations

  • Wang Xuan Institute of Computer Technology, Peking UniversityCore Member
    • Advisor: Prof. Xinhui Han
    • Research Focus: AI-empowered system security, intelligent attack and defense strategies, and adversarial attack research.
  • VUL337 Group, Tsinghua UniversityVisiting Master’s Student & Core Developer
    • Advisor: Prof. Chao Zhang
    • Key Contributions: Leading the indirect call (ICall) recognition project CALLEE and cross-architecture Binary Code Similarity Detection (BCSD); advancing controlled code generation and Machine Large Language Model (MLM) decompiler training; and developing software translation agents.

Education

  • Peking UniversityM.S. in Wang Xuan Institute of Computer Technology (2024 - Present)
  • Beijing University of Posts and Telecommunications (BUPT)B.S. in Automation (2020 - 2024)

Internship Experience

Alibaba | AAGI Research Intern

  • Core Responsibilities: Focused on Large Language Model (LLM) alignment, specifically designing Process Reward Models (PRM) and self-evolving mechanisms to enhance complex reasoning capabilities.
  • Technical Achievements: Engineered and deployed process-level reward models integrated with reinforcement learning algorithms (such as GRPO) to provide granular, step-by-step supervision for multi-turn reasoning and tool-integrated tasks.
  • Expertise: Developed a self-evolving framework that leverages automated verification and continuous feedback loops, enabling the model to autonomously iterate and optimize its reasoning policy without extensive human annotation.

ByteDance | Seed Talent Program - Seedance Intern

  • Core Responsibilities: Model training and architecture development(Infra).
  • Technical Achievements: Developed distributed training systems based on the Ray framework supporting large-scale training across thousands of GPUs. Gained extensive experience adapting and optimizing diverse heterogeneous computing platforms, including Huawei Ascend, H20, L20, and A100.
  • Expertise: Provided deep technical support and strategy iteration for the model post-training pipeline.

Tencent - Yuanbao | Post-Training Intern

  • Core Responsibilities: Focused on model alignment and reasoning capability enhancement during the LLM post-training phase.
  • Tech Stack: Applied cutting-edge algorithms such as Supervised Fine-Tuning and Reinforcement Learning alignment in production environments.

Huaqing Weiyang | Security Research Intern

  • Core Achievements: Led the fine-tuning research of Qwen-3 for binary security applications.
  • Academic Output: Proposed an innovative approach for cross-architecture code similarity detection. The research findings have been accepted by AAAI 2026.

Research Skills & Expertise

  • Binary Security: Cross-architecture function similarity detection, Indirect Call (ICall) recognition, and decompiler evaluation.
  • Large Language Models: RLHF/GRPO implementation, Process Reward Models (PRM), and tool-integrated reasoning.
  • System Architecture: Large-scale GPU cluster scheduling (Ray) and low-level Instruction Set Architecture (Cross-ISA) mapping analysis.