Xinyue XU

I am currently a final-year PhD student at The Hong Kong University of Science and Technology (HKUST), working under the supervision of Prof. Xiaomeng Li. Additionally, I work closely with Prof. Lijie Hu at MBZUAI and Prof. Sichao Li (Incoming Lecturer at USYD).

Prior to my PhD, I earned my Bachelor's degree (with Honours) from The Australian National University, where I was advised by Senior Prof. Amanda Barnard and supervised by Dr. Amanda Parker.

My research journey has been enriched by experiences at the MIT Computational Connectomics Group led by Prof. Nir Shavit, where I collaborated with Prof. Hao Wang and Prof. Lu Mi. I also gained valuable experience as a Research Assistant at the National University of Singapore, supervised by Dr. Gang Guo.

I am actively looking for collaborators in XAI and Healthcare. If you are interested, please reach out via e-mail.

Email Google Scholar Github

Publications & Manuscripts

(* indicates equal contribution, _ indicates equal advising)

Evaluating LLM Understanding via Structured Tabular Decision Simulations

Sichao Li*, Xinyue Xu*, Xiaomeng Li

Under Review, 2026

We introduce STaDS to evaluate LLM understanding beyond accuracy, revealing that models often lack faithful reliance on relevant decision factors despite their predictive success.

📄 Paper 💻 Code LLM Interpretability

Concept-Based Unsupervised Domain Adaptation

Xinyue Xu*, Yueying Hu*, Hui Tang, Yi Qin, Lu Mi, Hao Wang, Xiaomeng Li

ICML, 2025

We propose a framework - CUDA that improves interpretability and robustness of Concept Bottleneck Models under domain shifts through adversarial training.

📄 Paper 💻 Code Interpretability

Energy-Based Concept Bottleneck Models: Unifying Prediction, Concept Intervention, and Probabilistic Interpretations

Xinyue Xu, Yi Qin, Lu Mi, Hao Wang, Xiaomeng Li

ICLR, 2024

ECBM is a unified framework for concept-based prediction, concept correction, and fine-grained interpretations based on conditional probabilities.

📄 Paper 💻 Code Interpretability

Energy-Based Conceptual Diffusion Model

Yi Qin, Xinyue Xu, Hao Wang, Xiaomeng Li

NeurIPS Safe Generative AI Workshop, 2024

We introduce ECDMs, a unified energy-based framework that integrates concept-based generation, interpretation, debugging, and intervention.

📄 Paper Interpretability

Machine Learning Models for Evaluating Biological Reactivity within Molecular Fingerprints of Dissolved Organic Matter over time

Chen Zhao, Kai Wang, Qianji Jiao, Xinyue Xu, Yuanbi Yi, Penghui Li, Julian Merder, Ding He

Geophysical Research Letters, 2024

We utilize machine learning to correlate molecular composition with biological reactivity in reservoirs, leveraging model tuning to enhance both predictive power and interpretability.

AI4Science

Dynamic Data Augmentation via MCTS for Prostate MRI Segmentation

Xinyue Xu*, Yuhan Hsi*, Haonan Wang, Xiaomeng Li

ICONIP, 2023 (Oral Presentation)

We introduce Dynamic Data Augmentation (DDAug), a computationally efficient method to mitigate overfitting in medical image analysis caused by limited data availability.

📄 Paper 💻 Code AI4Science

Exploring the Complexities of Dissolved Organic Matter Photochemistry from the Molecular Level by Using Machine Learning Approaches

Chen Zhao*, Xinyue Xu*, et al.

Environmental Science & Technology (ES&T), 2023

The 6th Xiamen Symposium on Marine Environmental Sciences, 2023 (🏆 Best Poster Award)

We apply machine learning to integrate irradiation experiments, providing novel insights into the photochemical transformation processes of estuarine dissolved organic matter.

📄 Paper AI4Science

Subgraph Frequency Distribution Estimation using Graph Neural Network

Zhongren Chen*, Xinyue Xu*, Shengyi Jiang, Hao Wang, Lu Mi

KDD Deep Learning on Graphs, 2022

We propose GNNS, a graph neural network-based framework for efficient subgraph sampling and frequency distribution estimation in large networks.

📄 Paper AI4Science

Towards Faster Hyperparameter-free Clustering using Enhanced Iterative Label Spreading

Xinyue Xu

Honours Thesis, 2022

This work develops an enhanced, faster version of Iterative Label Spreading Clustering specifically optimized for materials science applications with reduced hyperparameters.

💻 Code AI4Science

Hybrid Model for Network Anomaly Detection with Gradient Boosting Decision Trees and Tabtransformer

Xinyue Xu, Xiaolu Zheng

ICASSP, 2021

We present our award-winning solution (Ranked 2nd) for the ICASSP 2021 Network Anomaly Detection Challenge.

📄 Paper Network Security

Classification Models for Medical Data with Interpretative Rules

Xinyue Xu, Xiang Ding, Zhenyue Qin, Yang Liu

ICONIP, 2021 (Oral Presentation)

We evaluate various classification models for SARS detection on diverse datasets, generating explanatory rules to ensure model interpretability.