Download CV

Education

• B.Eng. in Computer Science and Technology, Tsinghua University, Beijing, China, 2020-2024
• PhD candidate in Computer Science and Technology, Tsinghua University, Beijing, China, 2024-2029 (expected)

Academic

• GPA: 3.97 / 4.00, Rank: 4/181
• Courses of A+ : English for Academic Purposes (A): Spoken Communication, Probability and Statistics, Physics for Scientists and Engineers B(2), Artificial Neural Networks, Diploma Project(Thesis)
• Courses of A : Fundamentals of Computer Science, Fundamentals of Programming, Programing and Training, Foundation of Object-Oriented Programming, Assembly Language Programming, Calculus A(1), Calculus A(2), Linear Algebra, Advanced Topics in Linear Algebra, Discrete Mathematics(1), Physics for Scientists and Engineers B(1), Introduction to Artificial Intelligence, Software Engineering, Digital Logic Circuit, Digital Logic Experimentation, Fundamentals of Computer Graphics, Introduction to Modern Cryptography, Principles of Signal Processing, Principles and Practice of Compiler Construction, Introduction to High Performance Computing, Operating Systems, Database Special Topic Training, Cybersecurity Fundamentals
• A member of THUNLP (THU Natural Language Processing Group), advised by Associate Professor Zhiyuan Liu.

Publications

(*indicates equal contribution)

• Process Reinforcement through Implicit Rewards [Blog]
  Ganqu Cui*, Lifan Yuan*, Zefan Wang*, Hanbin Wang*, Wendi Li*, Bingxiang He*, Yuchen Fan*, Tianyu Yu*, Qixin Xu*, Weize Chen, Jiarui Yuan, Huayu Chen, Kaiyan Zhang, Xingtai Lv, Shuo Wang, Yuan Yao, Xu Han, Hao Peng, Yu Cheng, Zhiyuan Liu, Maosong Sun, Bowen Zhou, Ning Ding
  ICML 2025 Submission [code]

• EscapeBench: Pushing Language Models to Think Outside the Box [Paper]
  Cheng Qian, Peixuan Han, Qinyu Luo, Bingxiang He, Xiusi Chen, Yuji Zhang, Hongyi Du, Jiarui Yao, Xiaocheng Yang, Denghui Zhang, Yunzhu Li, Heng Ji
  ACL 2025 Submission [code]

• Zero-Shot Generalization during Instruction Tuning: Insights from Similarity and Granularity [Paper]
  Bingxiang He*, Ning Ding*, Cheng Qian*, Jia Deng, Ganqu Cui, Lifan Yuan, Haiwen Hong, Huan-ang Gao, Longtao Huang, Hui Xue, Huimin Chen, Zhiyuan Liu, Maosong Sun
  ACL 2025 Submission [code]

• Tell Me More! Towards Implicit User Intention Understanding of Language Model Driven Agents [Paper]
  Cheng Qian*, Bingxiang He*, Zhong Zhuang, Jia Deng, Yujia Qin, Xin Cong, Zhong Zhang, Jie Zhou, Yankai Lin, Zhiyuan Liu, Maosong Sun.
  ACL 2024 Main [code]

• UltraFeedback: Boosting Language Models with High-quality Feedback [Paper]
  Ganqu Cui*, Lifan Yuan*, Ning Ding, Guanming Yao, Bingxiang He, Wei Zhu, Yuan Ni, Guotong Xie, Ruobing Xie, Yankai Lin, Zhiyuan Liu, Maosong Sun.
  ICML 2024 Poster

• Beat LLMs at Their Own Game: Zero-Shot LLM-Generated Text Detection via Querying ChatGPT [Paper]
  Biru Zhu, Lifan Yuan, Ganqu Cui, Yangyi Chen, Chong Fu, Bingxiang He, Yangdong Deng, Zhiyuan Liu, Maosong Sun, Ming Gu.
  EMNLP 2023 Main [code]

• A Unified Evaluation of Textual Backdoor Learning: Frameworks and Benchmarks [Paper] (Spotlight)
  Ganqu Cui*, Lifan Yuan*, Bingxiang He, Yangyi Chen, Zhiyuan Liu, Maosong Sun.
  NeurIPS Datasets & Benchmarks 2022 [code]

Projects

• OpenBackdoor: An open-source toolkit for textual backdoor attack and defense [GitHub] [Paper]
  • Summarize three practical scenarios of attack methods based on their accessibility and goals.
  • Conclude novel metrics for three evaluation dimensions and recommend scenario-specified evaluation methodologies.
  • Develop an open-source toolkit OpenBackdoor and conduct extensive benchmark experiments.
  • Propose CUBE, a simple yet strong baseline method targeting purifying poisoned datasets.
• Tell Me More: Implementation and evaluation of Mistral-Interact [GitHub] [Paper]
  • Better understanding of user judgments
  • Comprehensive summarization of user intentions
  • Enhanced model-user interaction experience
  • Comparable performance with closed-source GPT-4
• UltraFeedback: A large-scale, fine-grained, diverse preference dataset [GitHub] [Paper]
  • Scale: UltraFeedback consists of 64k prompts, 256k responses and high-quality feedback. RLHF researchers could further construct around 340k comparison pairs to train their reward models.
  • Diversity: We collect prompts from various sources and query a diverse set of state-of-the-art open-source and prestigious models. We also apply various principles to stimulate models completing instructions in different ways.
  • High-density: UltraFeedback provides both numerical and textual feedback. Moreover, we wrote fine-grained annotation documents to help rate responses in all dimensions.
• PRIME: Scalable RL solution for advanced reasoning of language models [GitHub] [Blog]
  • We present PRIME (Process Reinforcement through IMplicit REwards), an open-source solution for online RL with process rewards, to advance reasoning abilities of language models beyond imitation or distillation.
  • With PRIME, starting from Qwen2.5-Math-7B-Base, our trained model Eurus-2-7B-PRIME achieves 26.7% pass@1 on AIME 2024, surpassing GPT-4o and Qwen2.5-Math-7B-Instruct. We achieve this with only 1/10 data of Qwen Math (230K SFT + 150K RL).
  • We also explore inference-time scaling and train EurusPRM, a SOTA-level math PRM that pushes the boundary even further.

Honors & Awards

• Second Prize in Freshmen Scholarship, Tsinghua University. (09, 2020)
• Second Prize in National Undergraduate Physics Competition, Beijing Physics Society. (04, 2021)
• Comprehensive Merit Scholarship of Tsinghua for the 2020-2021 school year, Dept. of CST. (10, 2021)
• Third Prize in THU Challenge Cup Academic Competition, Tsinghua University. (04, 2022)
• Comprehensive Merit Scholarship of Tsinghua for the 2021-2022 school year, Dept. of CST. (10, 2022)
• Comprehensive Merit Scholarship of Tsinghua for the 2022-2023 school year, Dept. of CST. (10, 2023)
• Five Star ZiJing Volunteer Award, Tsinghua University Communist Youth League Committee. (05, 2024)
• Outstanding Paper Award for Diploma Project, Tsinghua University. (06, 2024)
• Outstanding Graduate Award, Beijing Municipal Education Commission. (06, 2024)

Skills & Expertise

• Platform: Linux, Windows
• Tool: Git, SSH, Make, Tmux, Markdown, LaTeX
• Computer Language: C/C++ == Python > JavaScript == TypeScript > SQL
• Full-stack development: React, Django, Nginx, MySQL
• Rich experience in state-of-the-art deep learning techniques.