Cansu Sancaktar

I’m a PhD student working on intrinsically motivated RL and open-endedness, advised by Georg Martius. My work focuses on building agents that decide what’s worth learning and explore their environment efficiently and autonomously, much like children at play.

I'm excited about the continual acquisition of novel skills with minimal external supervision, progressing toward scalable, self-improving systems.

You can reach out to me at: firstname [dot] lastname [at] gmail [dot] com.

CV  /  Google Scholar  /  Twitter  /  Github

profile photo
Shadow Character

News
  • 05/2025 - Our paper "SENSEI: Semantic Exploration Guided by Foundation Models to Learn Versatile World Models" has been accepted to ICML 2025!
  • 04/2025 - Started my research internship at the CodeGen team at FAIR Meta Paris, advised by Taco Cohen.
  • 07/2024 - IMOL Workshop has been accepted to NeurIPS 2024 in Vancouver!
  • 06/2024 - Moving to Amsterdam for my internship at Qualcomm in the Embodied AI Team 🤖
  • show more
Publications & Preprints

Hybrid Training for Vision-Language-Action Models
Pietro Mazzaglia, Cansu Sancaktar, Markus Peschl, Daniel Dijkman
3rd CoRL Workshop on Learning Effective Abstractions for Planning, 2025
paper
Offline vs. Online Learning in Model-based RL: Lessons for Data Collection Strategies
Jiaqi Chen, Ji Shi, Cansu Sancaktar, Jonas Frey, Georg Martius
Reinforcement Learning Conference (RLC) 2025
paper

Modelling Microbial Communities with Graph Neural Networks
Albane Ruaud, Cansu Sancaktar, Marco Bagatella, Christoph Ratzke, Georg Martius
ICML 2024
paper / website
Optimistic Active Exploration of Dynamical Systems
Bhavya Sukhija, Lenart Treven, Cansu Sancaktar, Sebastian Blaes, Stelian Coros, Andreas Krause
NeurIPS 2023
paper / code
Real Robot Challenge 2022: Learning Dexterous Manipulation from Offline Data in the Real World
Nico Gürtler, Felix Widmaier, Cansu Sancaktar, ... [show all authors] , Georg Martius
NeurIPS 2022 Competition Track
paper

End-to-End Pixel-Based Deep Active Inference for Body Perception and Action
Cansu Sancaktar, Marcel A. J. Gerven, Pablo Lanillos
2020 Joint IEEE 10th International Conference on Development and Learning and Epigenetic Robotics (ICDL-EpiRob)
paper / video / code

Talks

11/2023 — Playful Exploration in Reinforcement Learning, BeNeRL Seminar Series.

Thanks Jon Barron for this nice template.