Yuejiang Liu

Welcome to my personal space on the internet.

I am Yuejiang Liu, a first-year Postdoc in the IRIS group at Stanford, advised by Chelsea Finn. My research is supported by an SNSF Postdoc Fellowship. I recently obtained my PhD degree from the VITA group at EPFL, advised by Alexandre Alahi, and previously spent time as a research intern in the CRL group, mentored by Francesco Locatello at ISTA, Chris Russell at Oxford, Bernhard Schölkopf at MPI-IS.

I enjoy designing and prototyping algorithms that enable deep neural networks to generalize robustly or adapt efficiently to new environments. In particular, I have been working on self-supervised learning, causal representation learning, and test-time adaptation, with applications in computer vision and multi-agent systems. Lately, I have been exploring these areas in the context of foundation models.

My long-term goal is to create embodied intelligent agents that can step out of lab demos and thrive in our ever-evolving world. If our research interests intersect, please don’t hesitate to get in touch :)

Publications

• Co-Supervised Learning: Improving Weak-to-Strong Generalization with Hierarchical Mixture of Experts.
  Y Liu, A Alahi.
  Preprint. 2024. [pdf, code]

• Sim-to-Real Causal Transfer: A Metric Learning Approach to Causally-Aware Interaction Representations.
  Y Liu*, A Rahimi*, P Luan*, F Rajic, A Alahi.
  In submission. Short version presented at NeurIPS workshops, 2023. [pdf, code]

• On Pitfalls of Test-time Adaptation.
  H Zhao*, Y Liu*, A Alahi, T Lin.
  International Conference on Machine Learning (ICML), 2023. [pdf, code]

• Causal Triplet: An Open Challenge for Intervention-centric Causal Representation Learning.
  Y Liu, A Alahi, C Russell, M Horn, D Zietlow, B Schölkopf, F Locatello.
  Conference on Causal Learning and Reasoning (CLeaR), 2023. [pdf, code]

• Towards Robust and Adaptive Motion Forecasting: A Causal Representation Perspective.
  Y Liu, R Cadei, J Schweizer, S Bahmani, A Alahi.
  Conference on Computer Vision and Pattern Recognition (CVPR), 2022. [pdf, code]

• Modular Low-Rank Style Transfer for Deep Motion Forecasting.
  P. Kothar*, D. Li*, Y Liu, A Alahi.
  Conference on Robot Learning (CoRL), 2022. [pdf, code]

• TTT++: When Does Self-supervised Test-time Training Fail or Thrive?
  Y Liu, P Kothari, B Delft, B Bellot-Gurlet, T Mordan, A Alahi.
  Conference on Neural Information Processing Systems (NeurIPS), 2021. [pdf, code]

• Social NCE: Contrastive Learning of Socially-aware Motion Representations.
  Y Liu, Q Yan, A Alahi.
  International Conference on Computer Vision (ICCV), 2021. [pdf, code]

• Collaborative Sampling from Generative Adversarial Networks.
  Y Liu*, P Kothari*, A Alahi.
  Conference on Artificial Intelligence (AAAI), 2020. [pdf, code]

• Crowd-Robot Interaction: Crowd-aware Robot Navigation with Attention-based Deep Reinforcement Learning.
  C Chen, Y Liu, S Kreiss, A Alahi.
  International Conference on Robotics and Automation (ICRA), 2019. [pdf, code]

• Map-based Deep Imitation Learning for Obstacle Avoidance.
  Y Liu, A Xu, Z Chen.
  International Conference on Intelligent Robots and Systems (IROS), 2018. [pdf]

• Real-Time Distributed Algorithms for Nonconvex Optimal Power Flow.
  Y Liu, JH Hours, G Stathopoulos, CN Jones.
  American Control Conference (ACC), 2017. [pdf]

Service

• Teaching Assistant: Deep Learning for Autonomous Vehicles (Guest Lecture on Self-supervised Learning)

• Review Service: NeurIPS, ICLR, ICML, CVPR, ICCV, ICRA, IROS, JMLR, PAMI, IJCV, RAL