Liu Junhao's Homepage

liujunhao@pku.edu.cn

Peking University

Beijing, China

Hello! Welcome to my homepage. I am Liusurname Junhaofirst name (刘俊豪/ljoʊ̯/ /tɕyn/ /xɑʊ̯/), a 4-th year PhD student in the Programming Language Lab at Peking University, advised by Prof. Zhang Xin.

My research interests lie in the field of eXplainable Artificial Intelligence (XAI), which focuses on developing machine learning and deep learning techniques that are transparent, interpretable, and understandable to humans. I am particularly interested in creating XAI methods that enhance human understanding and trust, and help people effectively utilize AI systems in real-world applications.

I was previously active in competitive programming, having earned several gold medals in the ICPC Asia-East Continental Final and Regional Contests. I also won a gold medal in the National Olympiad in Informatics (NOI) in 2017. Additionally, I am a co-founder of the PKU Student Algorithm Association.

In my spare time, I enjoy sim racing and am a fan of the Formula 1 racing series — especially the Scuderia Ferrari team (tifosi). I’m hoping to see them win the championship in ~~2025~~ 2026.

The most important lesson learned in programming language lab is to learn more languages. 除咗英文同普通話，我亦學緊粵語，咁我哋都可以用一齊練習粵語。

news

Mar 19, 2026	A new paper is available on arXiv: WASD: Locating Critical Neurons as Sufficient Conditions for Explaining and Controlling LLM Behavior.
Feb 12, 2026	A new paper is available on arXiv: Focus-LIME: Surgical Interpretation of Long-Context Large Language Models via Proxy-Based Neighborhood Selection.
May 20, 2025	A new paper is available on arXiv: Revitalizing Black-Box Interpretability: Actionable Interpretability for LLMs via Proxy Models.
Feb 01, 2025	A new paper is available on arXiv: MAnchors: Memorization-Based Acceleration of Anchors via Rule Reuse and Transformation.
Dec 13, 2024	Paper accepted at AAAI-25 (Oral): ReX: A framework for incorporating temporal information in model-agnostic local explanation techniques.

selected publications

AAAI

ReX: A framework for incorporating temporal information in model-agnostic local explanation techniques

Junhao Liu, and Xin Zhang

In Proceedings of the AAAI Conference on Artificial Intelligence, 2025

Oral Abs PDF

Oral (4.68%)

Existing local model-agnostic explanation techniques are ineffective for machine learning models that consider inputs of variable lengths, as they do not consider temporal information embedded in these models. To address this limitation, we propose ReX, a general framework for incorporating temporal information in these techniques. Our key insight is that these techniques typically learn a model surrogate by sampling model inputs and outputs, and we can incorporate temporal information in a uniform way by only changing the sampling process and the surrogate features. We instantiate our approach on three popular explanation techniques: Anchors, LIME, and Kernel SHAP. To evaluate the effectiveness of ReX, we apply our approach to six models in three different tasks. Our evaluation results demonstrate that our approach 1) significantly improves the fidelity of explanations, making model-agnostic techniques outperform a state-of-the-art model-specific technique on its target model, and 2) helps end users better understand the models’ behaviors.