Mert Albaba

bio

I am a PhD student at ETH Zürich and the Max Planck Institute – Perceiving Systems, co-advised by Michael Black and Andreas Krause. My primary research interests span reinforcement learning, imitation learning, robotics, diffusion models and VLMs. I am especially excited about creating generative ways that enable humanoid agents to learn wide range of skills autonomously. I am an ELLIS PhD Student and a CLS Fellow.

I previously spent a year at USC, earned my MSc in Mechanical Engineering at Bilkent University under the supervision of Yildiray Yildiz, and completed my BSc in Electrical and Computer Engineering at Bilkent University.

Download my CV

research

How can we enable robots and physical characters autonomously acquire new skills? I develop multimodal imitation learning approaches that combine (inverse) reinforcement learning, vision language models (VLMs), and pretrained generative models. Ultimately, I aim to enable humanoid agents, virtual or physical, to master a wide range of tasks by learning from just a handful of demonstrations or even from generated data alone. I’m particularly interested in both robotics and realistic avatar animation (e.g., character simulation).

In addition to my work in generative imitation learning, I have 7+ years of experience in reinforcement learning and imitation learning to model human behavior in multi-agent settings. I also remain interested in multi-agent settings, where game-theoretic principles can enhance both reinforcement and imitation learning.