Tewodros W. Ayalew

PhD Student in Computer Science at the University of Chicago

prof_pic.jpg

Email: tewodrosayalew [at] uchicago [dot] edu

I’m a PhD student in Computer Science at the University of Chicago, advised by Dr. Matthew Walter at the Robot Intelligence through Perception Laboratory (RIPL) at the Toyota Technological Institute at Chicago (TTIC).

I’m interested in how robots can learn about the physical world from video. My research focuses on world models, learning latent actions and dynamics from unlabeled videos, and learning reward signals from visual observations, with the goal of agents that learn from diverse real-world experience and transfer across tasks and embodiments.

Before my PhD, I received an M.Sc. in Computer Science from the University of Saskatchewan, where I worked with Dr. Ian Stavness on unsupervised domain adaptation.

news

Oct 01, 2026 Our paper “What Are We Actually Benchmarking in Robot Manipulation?” was accepted as an oral at CoRL 2026, and received the Best Abstract Award at the IROS 2026 Workshop on Reproducible Benchmarking of Robotic Grasping and Manipulation! 🎉
Aug 31, 2026 Our paper “CLAW: Learning Continuous Latent Action World Models via Adversarial Latent Regularization” was selected for a lightning talk at WM@Booth 2026! 🎉
Jun 11, 2026 I gave a talk on “CLAW: Learning Continuous Latent Action World Models via Adversarial Latent Regularization” at the Midwest Robotics Workshop (MWRW) 2026! 🎉
Jun 26, 2025 Our paper “PROGRESSOR: A Perceptually Guided Reward Estimator with Self-Supervised Online Refinement” is accepted at ICCV 2025! 🎉 😊
Feb 10, 2025 Our paper “Enabling End Users to Program Robots Using Reinforcement Learning” has been accepted at HRI 2025.
Nov 26, 2024 Our paper “PROGRESSOR: A Perceptually Guided Reward Estimator with Self-Supervised Online Refinement” is now available on Arxiv.
Jun 15, 2021 Our paper “Automatic Microplot Localization Using UAV Images and a Hierarchical Image-based Optimization Method” has been published in Plant Phenomics.
Aug 28, 2020 Our paper “Unsupervised Domain Adaptation for Plant Organ Counting” received the Best Paper Award in CVPPP at ECCV 2020! 🎉 😊
Aug 20, 2020 Our paper “Unsupervised Domain Adaptation for Plant Organ Counting” has been accepted at ECCV 2020.
Aug 15, 2020 Our paper “AutoCount: Unsupervised Segmentation and Counting of Organs in Field Images” has been accepted at ECCV 2020.

selected publications

  1. WM@Booth
    CLAW: Learning Continuous Latent Action World Models via Adversarial Latent Regularization
    Tewodros W. Ayalew, Matthew Jeung, Samuel Wheeler, and 5 more authors
    In Workshop on World Models at Chicago Booth (WM@Booth), 2026
    Takeaway: end-to-end continuous latent actions + diffusion world model from action-free video; 43% RoboDesk upright-block planning vs. 5% AdaWorld; 1.7/3.0 long-horizon real UR5 vs. 1.5 AdaWorld.
    Lightning Abs arXiv Project Page PDF
  2. ICCV
    PROGRESSOR.gif
    PROGRESSOR: A Perceptually Guided Reward Estimator with Self-Supervised Online Refinement
    Tewodros W. Ayalew, Xiao Zhang, Kevin Yuanbo Wu, and 3 more authors
    In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2025
    Takeaway: self-supervised visual progress reward with online push-back; Meta-World RL without env reward; real-UR5 few-shot offline RL under noisy demos.
    Abs Project Page Code
  3. CoRL
    What Are We Actually Benchmarking in Robot Manipulation?
    Tianchong Jiang, Xiangshan Tan, Samuel Wheeler, and 3 more authors
    In Proceedings of the Conference on Robot Learning (CoRL), 2026
    Takeaway: four diagnostics showing popular manipulation benchmarks often fail as proxies for general capability; reference implementations released.
    Oral Abs arXiv Project Page PDF Code
  4. uda4poc.gif
    Unsupervised Domain Adaptation for Plant Organ Counting
    Tewodros W. Ayalew, Jordan Ubbens, and Ian Stavness
    In Proceedings of the European Conference on Computer Vision (ECCV), 2020
    Best Paper Project Page