Can Mizrakli

MSc Computer Science at Karlsruhe Institute of Technology

can-mizrakli.jpeg

I work on robot learning, with a focus on vision-language-action models and long-horizon manipulation.

I’m a research assistant at the CV:HCI Lab, working in collaboration with the H²T Lab under the supervision of Dr.-Ing. David Schneider, on adapting VLA models to new robot embodiments such as the ARMAR-6/7 humanoids and bimanual Franka platforms.

Previously, over the 2025–26 academic year, I worked with the Intuitive Robots Lab under Prof. Rudolf Lioutikov on memory-based imitation learning, as part of a university Seminar and Praktikum. Our paper ReCAT is available as an arXiv preprint, and its short version was presented as a poster at the CoRL 2026 Memory for Robot Foundation Models workshop.

Research interests

  • Robot learning, imitation learning, and robot manipulation
  • Vision-language-action foundation models
  • State space models and memory for long-horizon control
  • Reinforcement learning for robotics

Read more about my research →

News

Nov 12, 2026 ReCAT accepted as a poster at the CoRL 2026 Memory for Robot Foundation Models workshop in Austin! 🎉 [PDF]
Sep 28, 2026 ReCAT: Remember, Count, And Time preprint is out on arXiv! 🆕
Jul 14, 2026 TDSal: Task-Based Top-Down Saliency Prediction Model preprint is out on arXiv!

Publications

  1. recat_banner.jpg
    ReCAT: Remember, Count, And Time: Structured Recurrent Memory for Robot Manipulation
    P. Vanjani*, M. Hatab*, C. Mizrakli*, V. Shaj, Z. Li, M. Reuss, and R. Lioutikov
    arXiv preprint arXiv:2609.35200, Sep 2026
  2. tdsal.jpg
    TDSal: Task-Based Top-Down Saliency Prediction Model
    Can Mizrakli and Tolga K Capin
    In CGI 2026, LNCS Proceedings (to appear), Jul 2026