About me

Hello! I am Yuxuan Zhu, a third-year PhD student in Computer Science at Rensselaer Polytechnic Institute (RPI), advised by Prof. Mohammad Mohammadi Amiri. I received a bachelor’s degree in Electrical Engineering (Medical Informatics track) from Sichuan University in China and a master’s degree in Computer Science from Leiden University in the Netherlands, where I was advised by Prof. Matthijs van Leeuwen and Dr. Zhong Li.

I am looking for research internship in LLM efficiency (inference, reasoning, agentic methods, etc.). I am also open to research collaboration opportunities.

Research Interests

My research interests lie in efficient and reliable machine learning, with a current primary focus on Large Language Models (LLMs). I am particularly interested in making LLM inference more memory and computation efficient, with recent work on KV cache compression, adaptive memory mechanisms, and reasoning-aware generation. I am also interested in improving the efficiency of multimodal models and agentic AI systems.

I also have research experience in Graph Neural Networks (GNNs), particularly in robustness, security, and interpretability, as well as broader experience in Explainable AI (XAI) and anomaly detection.

News

• [03/2026] I will join IBM Research for an research internship this summer.
• [09/2025] One paper about Low-Rank KV Cache Compression is now available on arXiv. [paper]
  
• [09/2025] Our paper titled Scalable, Explainable and Provably Robust Anomaly Detection with One-Step Flow Matching was accepted by NeurIPS 2025 for publication!
  
• [07/2025] Our paper titled SentenceKV: Efficient LLM Inference via Sentence-Level Semantic KV Caching was accepted by Conference on Language Modeling (COLM) 2025 for publication! [code]
  
• [08/2024] Our paper titled On the Robustness of Graph Reduction Against GNN Backdoor was accepted by AISec@CCS 2024 for publication!
  
• [09/2023] Moved from 🇳🇱 to 🇺🇸. Started my PhD journey at RPI!
  
• [07/2023] Our paper titled A Survey on Explainable Anomaly Detection was accepted by TKDD for publication!
  
• [03/2023] Graduated from Leiden University in the Netherlands with a master’s degree in Computer Science [thesis].
  
• [02/2023] Completed my graduation thesis/internship Data-driven Nozzle Failures Detection and Classification at Canon Production Printing.

Services

• Reviewer, Conference on Neural Information Processing Systems (NeurIPS), 2024,2025
  
• Reviewer, International Conference on Distributed Computing Systems (ICDCS), 2024
  
• Reviewer, Transactions on Information Forensics & Security
  
• Reviewer, The International Conference on Learning Representations (ICLR), 2025,2026
  
• Reviewer, Empirical Methods in Natural Language Processing (EMNLP), 2025

Awards

• Conference on Language Modeling (COLM) 2025 Student Travel Grant.