Xuemin Yu

About me

I am a Ph.D. student in Computer Science at Dalhousie University, supervised by Dr. Hassan Sajjad. I am a member of the Hypermatrix Lab, and my research is in natural language processing and explainable AI.

I work on methods that make model predictions and internal representations easier to inspect, with recent work on latent concept-based explanations, scalable concept discovery, and cross-layer interpretation of language models.

Research Interests

My research focuses on interpretable and trustworthy NLP systems. Current topics include:

  • Concept-based Interpretability: Developing explanations that connect language-model predictions to latent concepts in representation space.
  • Latent Concept Discovery: Studying discrete and cross-layer concept discovery methods for interpreting encoder- and decoder-based language models.
  • Scalable Explanation Methods: Designing efficient alternatives to clustering-based concept discovery for large-scale hidden-state analysis.
  • Visual Analytics: Building interactive systems that help people inspect text-derived signals and sentiment patterns.

News

  • Vector Quantized Latent Concepts was updated on arXiv.
  • Static Sandboxes Are Inadequate was released on arXiv.
  • Cross-Layer Discrete Concept Discovery was released on arXiv.
  • Latent Concept-based Explanation of NLP Models appeared at EMNLP 2024.
  • Senti-COVID19 was published in IEEE Access.

Publications

2026

Vector Quantized Latent Concepts: A Scalable Alternative to Clustering-Based Concept Discovery

Xuemin Yu, Ankur Garg, Samira Ebrahimi Kahou, and Hassan Sajjad

arXiv. arXiv preprint arXiv:2602.02726.

arXivPDFDOI
2025

Static Sandboxes Are Inadequate: Modeling Societal Complexity Requires Open-Ended Co-Evolution in LLM-Based Multi-Agent Simulations

Jinkun Chen, Sher Badshah, Xuemin Yu, and Sijia Han

arXiv. arXiv preprint arXiv:2510.13982.

arXivPDFDOI
2025

Cross-Layer Discrete Concept Discovery for Interpreting Language Models

Ankur Garg, Xuemin Yu, Hassan Sajjad, and Samira Ebrahimi Kahou

arXiv. arXiv preprint arXiv:2506.20040.

arXivPDFDOI
2024

Latent Concept-based Explanation of NLP Models

Xuemin Yu, Fahim Dalvi, Nadir Durrani, Marzia Nouri, and Hassan Sajjad

EMNLP 2024. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 12435-12459.

ACLPDFDOI
2021

Senti-COVID19: An Interactive Visual Analytics System for Detecting Public Sentiment and Insights Regarding COVID-19 From Social Media

Xuemin Yu, Martha Dais Ferreira, and Fernando V. Paulovich

IEEE Access. IEEE Access, volume 9, pages 126684-126697.

DOI

Education

Ph.D. in Computer Science

Dalhousie University, Faculty of Computer Science

Supervisor: Dr. Hassan Sajjad

Bachelor of Computer Science with First Class Honours

Dalhousie University, Faculty of Computer Science

Honours supervisor: Dr. Fernando Paulovich; reader: Dr. Evangelos E. Milios

Teaching

  • Teaching Assistant, CSCI1108 Introduction to Experimental Robotics, Fall 2024
  • Teaching Assistant, CSCI3160 Designing User Interfaces, Winter 2024
  • Teaching Assistant, CSCI3160 Designing User Interfaces, Fall 2023
  • Teaching Assistant, CSCI3160 Designing User Interfaces, Winter 2023

Work

Part-time Instructor

Orange Education Ltd., Halifax, NS, Canada

Prepared lesson plans and instructed students in computer science-related courses.

Software Developer

Exorcat Technologies Ltd., Halifax, NS, Canada

Contributed to team-based software and web application development.