Publications

publications by categories in reversed chronological order. generated by jekyll-scholar.

2025

  1. ICDCS
    HybridServe: Efficient Serving of Large AI Models with Confidence-Based Cascade Routings
    Leyang Xue, Yao Fu, Luo Mai, and Mahesh Marina
    In ICDCS (In Conjunction Events), 2025
  2. ICDCS
    TUBO: A Tailored ML Framework for Reliable Network Traffic Forecasting
    Zhihang Yuan, Leyang Xue, Waleed Ahsan, and Mahesh Marina
    In ICDCS (In Conjunction Events), 2025
  3. arXiv
    MoE-Gen: High-Throughput MoE Inference on a Single GPU with Module-Based Batching
    Tairan Xu, Leyang Xue, Zhan Lu, and Luo Mai
    2025
    arXiv Code
  4. NSDI
    Towards Energy Efficient 5G vRAN Servers
    Anuj Kalia, Nikita Lazarev, Leyang Xue, Xenofon Foukas, Bozidar Radunovic, and Francis Y. Yan
    In NSDI, 2025

2024

  1. arXiv
    MoE-Infinity: Offloading-Efficient MoE Model Serving
    Leyang Xue, Yao Fu, Zhan Lu, Luo Mai, and Mahesh K. Marina
    2024
    arXiv Code
  2. arXiv
    MoE-CAP: Cost-Accuracy-Performance Benchmarking for Mixture-of-Experts Systems
    Yao Fu, Yinsicheng Jiang, Yeqi Huang, Ping Nie, Zhan Lu, Leyang Xue, Congjie He, Man-Kit Sit, and 5 more authors
    2024
    arXiv Code
  3. OSDI
    ServerlessLLM: Low-Latency Serverless Inference for Large Language Models
    Yao Fu, Leyang Xue, Yeqi Huang, Andrei-Octavian Brabete, Dmitrii Ustiugov, Yuvraj Patel, and Luo Mai
    In OSDI, 2024
    Code

2022

  1. ICNP
    PAINT: Path Aware Iterative Network Tomography for Link Metric Inference
    Leyang Xue, Mahesh K. Marina, Geng Li, and Kai Zheng
    In ICNP, 2022