publications

* denotes equal contribution. More to come:)

2026

2026

  1. TMLR
    ReDiTT: Retrieval Augmented Conditional Diffusion Transformers for Asynchronous Time Series
    Saiyue Lyu, Zhitian Zhang, Ruizhi Deng, and Thibaut Durand
    Transactions on Machine Learning Research, 2026
    PDF Code Poster ICLR Workshop Version
  2. ICML WS
    When Does Diffusion Purification Amplify Perturbations?
    Haibo Zhang, Saiyue Lyu, Yuntao Wang, and Takeshi Saitoh
    Foundations of Deep Generative Models (FoGen) Workshop at ICML, 2026
    PDF
  3. ISSAC
    Fast Decomposition of Sparse Polynomials
    Mark Giesbrecht, Pascal Koiran, Saiyue Lyu, and Daniel Roche (α-β)
    In Proceedings of the ACM International Symposium on Symbolic and Algebraic Computation, 2026
    PDF Master Thesis Version

2025

2025

  1. ICML WSOral
    Adaptive Diffusion Denoised Smoothing: Certified Robustness via Randomized Smoothing with Differentially Private Guided Denoising Diffusion
    Frederick Shpilevskiy, Saiyue Lyu, Krishnamurthy Dj Dvijotham, Mathias Lécuyer, and Pierre-André Noël
    2nd Test-Time Adaptation (PUT) Workshop at ICML, 2025
    PDF Poster

2024

2024

  1. NeurIPSSpotlight
    Adaptive Randomized Smoothing: Certified Adversarial Robustness for Multi-Step Defences
    Saiyue Lyu *, Shadab Shaikh *, Frederick Shpilevskiy *, Evan Shelhamer, and Mathias Lécuyer
    Advances in Neural Information Processing Systems, 2024
    Spotlight Highlighted [Top 2% in 15671]
    PDF Code Poster Slides CVPR Workshop Version
  2. TMLR
    DP-LDMs: Differentially Private Latent Diffusion Models
    Michael F Liu *, Saiyue Lyu *, Margarita Vinaroz *, and Mijung Park
    Transactions on Machine Learning Research, 2024
    PDF Code ICLR Workshop Version

2023

2023

  1. CSCW
    Exploring Temporal and Multilingual Dynamics of Post-Disaster Social Media Discourse: A Case of Fukushima Daiichi Nuclear Accident
    Saiyue Lyu, and Zhicong Lu
    In Proceedings of the ACM on Human-Computer Interaction (CSCW), 2023
    PDF