Xinyu Guo

I am a second-year Ph.D. student in Computer Science at the University of Arizona. I am fortunate to be advised by Prof. Mihai Surdeanu. Previously, I earned my B.E. degree in Computer Science from Shandong University in 2023.

My research focuses on Natural Language Processing (NLP), Reinforcement Learning (RL), and Explainable Large Language Models (LLMs). I am particularly interested in developing interpretable and transparent systems that can articulate, in human-understandable terms, the reasoning processes behind their decisions, thereby enhancing reliability, accountability, and trust of AI systems.

selected publications

ArXiv

Peeking inside the Black-Box: Reinforcement Learning for Explainable and Accurate Relation Extraction

Xinyu Guo, Zhengliang Shi, Minglai Yang, and 2 more authors

2026

In Submission to ICLR

ARXIV Bib PDF Code

@misc{guo2025peekinginsideblackboxreinforcement,
  title = {Peeking inside the Black-Box: Reinforcement Learning for Explainable and Accurate Relation Extraction},
  author = {Guo, Xinyu and Shi, Zhengliang and Yang, Minglai and Rahimi, Mahdi and Surdeanu, Mihai},
  html_author = {<a href="https://xinnyuguo.github.io">Xinyu Guo</a>, 
                    <a href="https://shizhl.github.io">Zhengliang Shi</a>, 
                    <a href="https://ymingl.com">Minglai Yang</a>, 
                    <a href="https://surdeanu.cs.arizona.edu/mihai/">Mihai Surdeanu</a>},
  year = {2026},
  note = {In Submission to ICLR},
  eprint = {2510.06198},
  archiveprefix = {arXiv},
  primaryclass = {cs.CL},
}

ArXiv

AlignSAE: Concept-Aligned Sparse Autoencoders

Minglai Yang, Xinyu Guo, Mihai Surdeanu, and 1 more author

2025

ARXIV