BeNeRL Reinforcement Learning Seminar: Michal Nauman from the University of Warsaw (July 9, this Thursday)

[Zhao Yang]

Dear colleagues,

Our next BeNeRL Reinforcement Learning Seminar (July 9) is coming: 

Speaker: Michal Nauman (https://scholar.google.com/citations?user=GnEVRtQAAAAJ&hl=en), PhD student from the University of Warsaw.

Title: Scaling in value-based RL

Date: July 9, 16.00-17.00 (Amsterdam time zone)

Please find full details about the talk below this email and on the website of the seminar series: https://www.benerl.org/seminar-series

The goal of the online BeNeRL seminar series is to invite RL researchers (mostly advanced PhD or early postgraduate) to share their work. In addition, we invite the speakers to briefly share their experience with large-scale deep RL experiments, and their style/approach to get these to work.

 

We would be very glad if you forward this invitation within your group and to other colleagues that would be interested (also outside the BeNeRL region). Hope to see you on July 9!

Kind regards,

Zhao Yang & Thomas Moerland

VU Amsterdam & Leiden University

——————————————————————

Upcoming talk:

Date: July 9, 16.00-17.00 (Amsterdam time zone)

Speaker: Michal Nauman

Title: ​Scaling in value-based RL

Zoom: https://universiteitleiden.zoom.us/j/65411016557?pwd=MzlqcVhzVzUyZlJKTEE0Nk5uQkpEUT09

Abstract: This talk will discuss compute scaling in value-based reinforcement learning, with a focus on scaling the number of parameters in value networks trained with temporal-difference learning. I will first discuss why naively scaling the critic can be difficult, and highlight the failure modes and symptoms that arise in practice. I will then present techniques that remedy these issues and enable effective critic scaling, leading to substantial performance gains over traditional architectures. Finally, I will show how scaled critic networks address challenges in multi-task RL, enabling highly efficient learning across diverse tasks.

Bio: Michal Nauman defended his PhD in 2026 under the supervision of Marek Cygan, with a dissertation titled "Sample-Efficient Actor-Critic Algorithms in Reinforcement Learning". During his PhD, he was a visiting researcher at UC Berkeley, where he worked with Pieter Abbeel, and spent one year at IDEAS NCBR, an ELLIS-affiliated research institute. His research focuses on value learning, scaling value-based reinforcement learning, and multi-task learning.