Lejs Deen Behric | ETH Zürich

About Me

I am a Master’s student in Computer Science at ETH Zürich where I also completed my Bachelor’s degree (GPA 5.85/6.0).

Research Interests

My research interests are in Machine Learning, with a focus on optimization and learning algorithms for large-scale models.

News

[Jun 2026] Started working as a Member of Technical Staff Intern at Amazon AGI Labs in San Francisco, working on RL post-training for agents.
[Apr 2026] Reinforcement Learning via Self-Distillation received the Best Paper Award at the ICLR 2026 Workshop on Test-Time Updates and was accepted to ICML 2026.
[Mar 2026] Reinforcement Learning via Self-Distillation received a spotlight at the RSI workshop and an oral at the TTU workshop, ICLR 2026!
[Jan 2026] Started a research internship in autonomous driving at Mercedes-Benz AG, Sindelfingen, Germany.
[Jan 2026] Our paper ConMeZO: Adaptive Descent-Direction Sampling for Gradient-Free Finetuning of Large Language Models was accepted to AISTATS 2026!
[Dec 2025] Preprint of Reinforcement Learning via Self-Distillation is out!

Publications

Reinforcement Learning via Self-Distillation

Jonas Hübotter, Frederike Lübeck*, Lejs Behric*, Anton Baumann*, Marco Bagatella, Daniel Marta, Ido Hakimi, Idan Shenfeld, Thomas Kleine Buening, Carlos Guestrin, Andreas Krause

International Conference on Machine Learning (ICML), 2026. Best Paper Award at ICLR 2026 Workshop on Test-Time Updates and Oral Presentation at ICLR 2026 Workshop on Scaling Post-training for LLMs.

PDF Code Project Page
ConMeZO: Adaptive Descent-Direction Sampling for Gradient-Free Finetuning of Large Language Models

Lejs Deen Behric, Liang Zhang, Bingcong Li, Kiran Koshy Thekumparampil

International Conference on Artificial Intelligence and Statistics (AISTATS), 2026.

PDF Code

Teaching

Teaching Assistant at Department of Computer Science, ETH Zürich, Switzerland (Sep 2023 – Jul 2025)

Introduction to Machine Learning (Prof. Andreas Krause, Prof. Fanny Yang), Spring 25
Data Structures and Algorithms (Prof. David Steurer, Prof. Johannes Lengler), Fall 23–25
Digital Design and Computer Architecture (Prof. L. Gürkaynak, Prof. M. Sadrosadati), Spring 24

Responsibilities included leading weekly sessions for classes of around 30 students, supporting and teaching students during exercises and labs, proofreading and refining lecture materials, contributing to the writing of course scripts, and assisting with exam supervision, grading, and feedback.

Talks

Reinforcement Learning via Self-Distillation, Amazon AGI Lab Research Reading Group, Apr 2026.