Scholar
Ruixuan Tu
Google Scholar ID: V6hzHoQAAAAJ
Department of Computer Science, The University of Tokyo
Citations & Impact
Citations
67
H-index
3
i10-index
2
Publications
10
Co-authors
0
Publications
Resume (English only)
Academic Achievements
- Published Papers: 'Is Semantic Chunking Worth the Computational Cost?' (Renyi Qu, Ruixuan Tu, Forrest Sheng Bao); 'FaithBench: A Diverse Hallucination Benchmark for Summarization by Modern LLMs' (Forrest Sheng Bao et al.).
Research Experience
- Worked on multiple projects to align LLMs with human expectations and behaviors, including: reference-free evaluation of summary, hallucination detection of summary, and retrieval-augmented generation through semantic chunking. Ongoing research includes: generalizability of in-context learning for transformer models and its applications in pluralistic alignment (with Prof. Ramya Korlakai Vinayak); opinion dynamics alignment in LLM agents (with Prof. Junjie Hu and Dr. Yun-Shiuan (Sean) Chuang).
Education
- The University of Tokyo, Master's in Computer Science, Advisor: Prof. Hitomi Yanaka, Expected Graduation: September 2027; University of Wisconsin--Madison, Undergraduate in Computer Sciences (Honors), Mathematics (Honors), Data Science, Statistics, and Japanese, Advisors: Prof. Forrest Sheng Bao (Iowa State CS and ML Head @ Vectara Inc), Prof. Ramya Korlakai Vinayak (UW--Madison ECE & CS & Stat), and Prof. Junjie Hu (UW--Madison BMI & CS), Expected Graduation: September 2027.
Background
- Research Interests: Human-aligned Large Language Models (LLMs), Multilingual NLP and Computational Linguistics (especially Japanese NLP). Major: Computer Sciences (Honors), Mathematics (Honors), Data Science, Statistics, and Japanese. Brief Introduction: Currently a fourth-year undergraduate at the University of Wisconsin--Madison, majoring in Computer Sciences (Honors), Mathematics (Honors), Data Science, Statistics, and Japanese. Will be joining The University of Tokyo as a master's student in Computer Science, researching in LLMs, starting from October 2025.
Miscellany
- Personal Interests: Connecting NLP knowledge with Japanese linguistics and classical Japanese courses, applying multilingual transfer learning from modern Japanese to classical Japanese, and using computational linguistics tools to analyze the morpheme origins in Japanese literature.
Co-authors