I am Ruomin Huang (黄若民 in Chinese), a 2nd-year CS Ph.D. student at Duke University. I am very fortunate to be advised by Prof. Rong Ge. Previously I received M.S. in Data Science and B.S. in Computational Mathematics from USTC, where I worked with Prof. Hu Ding.

Research

Previously I worked on the algorithmic aspect of ML. Now I am interested in mechanisms physics of transformers.

Publications

* denotes equal contribution.

• Task Descriptors Help Transformers Learn Linear Models In-Context
  Ruomin Huang and Rong Ge
  ICLR 2025
  Short version appears at 1st ICML Workshop on In-Context Learning (ICL@ICML 2024)

• To Tackle Adversarial Transferability: A Novel Ensemble Training Method with Fourier Transformation
  Wanlin Zhang, Weichen Lin, Ruomin Huang, Shihong Song and Hu Ding
  ICLR 2025

• ReCaLL: Membership Inference via Relative Conditional Log-Likelihoods
  Roy Xie, Junlin Wang, Ruomin Huang, Minxing Zhang, Rong Ge, Jian Pei, Neil Gong and Bhuwan Dhingra
  EMNLP 2024

• An Effective Dynamic Gradient Calibration Method for Continual Learning
  Weichen Lin, Jiaxiang Chen, Ruomin Huang and Hu Ding
  ICML 2024

• Coresets for Wasserstein Distributionally Robust Optimization Problems
  Ruomin Huang, Jiawei Huang, Wenjie Liu and Hu Ding
  NeurIPS 2022 (spotlight)

• Coresets for Relational Data and The Applications
  Jiaxiang Chen, Qingyuan Yang, Ruomin Huang and Hu Ding
  NeurIPS 2022 (spotlight)

• A Novel Sequential Coreset Method for Gradient Descent Algorithms
  Jiawei Huang*, Ruomin Huang*, Wenjie Liu*, Nikolaos M. Freris and Hu Ding
  ICML 2021 (spotlight)

• Randomized Greedy Algorithms and Composable Coreset for k-Center Clustering with Outliers
  Hu Ding, Ruomin Huang, Kai Liu, Haikuo Yu and Zixiu Wang
  Preprint

Teaching Assistant

Undergraduate course

• 01118601 Algorithms for Big Data (2021 spring, USTC)
• COMPSCI 330 Design and Analysis of Algorithms (2024 spring, Duke)

Doctoral course

• COMP7102P01 Advanced Algorithms Design and Analysis (2021 fall, USTC)

Services

• Reviewer for ICML, NeurIPS, ICLR and SoCG.