Publications

Journal & Conference Papers

• Xie, S., Yuan, R., Rossi, S., Hannagan, T.
  The Initialization Determines Whether In‑Context Learning Is Gradient Descent.
  Transactions on Machine Learning Research (TMLR), 2025.

• Xie, S., Redko, I., Feofanov, V., Alonso, M., Odonnat, A., Zhang, J., Palpanas, T.
  CauKer: Classification Time Series Foundation Models Can Be Pretrained on Synthetic Data Only.
  ICML 2025 Workshop on Foundation Models for Structured Data (Best Time Series Paper), 2025.
  arXiv

• Xie, S., Aref,E., Giraldo, J. H.
  Subgraph Gaussian Embedding Contrast for Self‑Supervised Graph Representation Learning.
  ECML‑PKDD 2025.

• Xie, S., Giraldo, J. H.
  Variational Graph Contrastive Learning.
  NeurIPS 2024 Workshop on Self‑Supervised Learning – Theory and Practice, 2024.
  View Paper

• Xie, S., Liu, Y., Shuai, W.
  FTUnet: Feature Transferred U‑Net for Single HDR Image Reconstruction.
  ACM Multimedia Asia (MMA), 2023 – Oral Presentation.
  View Paper

• Xie, S., Zhu, S.
  Feasibility Study of Intelligent Healthcare Based on Digital Twin and Data Mining.
  International Conference on Computer Information Science and Artificial Intelligence (CISAI), 2021.
  View Paper

Patents

• Neural Network Parameter Diffusion – Xie, S., Yuan, R., Rossi, S., Hannagan, T.
  Utilizes autoencoders and latent diffusion to compress and generate experts within mixture‑of‑experts models.

• Permutation Symmetries Applied to DeepSeek Mixture of Experts Language Models – Xie, S., Yuan, R., Rossi, S., Hannagan, T.
  Introduces a novel weight permutation symmetry for aligning experts in GLU‑based MoE architectures, enabling efficient compression, fine‑tuning and merging for models like DeepSeek‑MoE‑16B and Qwen1.5‑MoE‑A2.7B.

Reviewer & Service

• Reviewer for NeurIPS 2024 Workshop on Compression, COLM 2025, and NeurIPS 2025.