Scholar
Sophia Xiao Pu
Google Scholar ID: rRazhgkAAAAJ
PhD student, University of California, Santa Barbara
Citations & Impact
Citations
448
H-index
7
i10-index
4
Publications
11
Co-authors
4
list available
Publications
Resume (English only)
Academic Achievements
- Publications include:
- - EMNLP 2025: Dynamic Evaluation for Oversensitivity in LLMs
- - COLM 2025: THOUGHT TERMINATOR: Benchmarking, Calibrating, and Mitigating Overthinking in Reasoning Models
- - NAACL 2025 (Oral): B^4: A Black-Box Scrubbing Attack on LLM Watermarks
- Other academic activities include an internship at AWS, Santa Clara.
Research Experience
- Currently a second-year CS PhD student at the University of California, Santa Barbara, advised by Prof. William Wang. Research projects include trustworthy AI, efficiency, and other related areas.
Education
- Bachelor’s degree from Peking University, advised by Prof. Xiaojun Wan. Also worked with Prof. Tianxing He at Tsinghua University and Prof. Yulia Tsvetkov at the University of Washington.
Background
- Research interests lie broadly in Language and Vision, particularly in: Trustworthy AI (detecting machine-generated text, attacking LLM watermarks, evaluating oversensitivity in LLMs), Efficiency (prompt compression, overthinking in reasoning models), and Other (extrinsic evaluation for text summaries). Currently a second-year CS PhD student at the University of California, Santa Barbara.
Miscellany
- Actively looking for motivated undergraduate or master’s students to collaborate on exciting topics such as multimodal evaluation, reasoning, and more.
