Vincent Brockers

I obtained my B.Sc. and M.Sc. degrees in Physics from the University of Göttingen. In my bachelor’s thesis, I investigated information spreading in networks with different topologies. During my master’s degree, I specialized in the physics of complex systems and completed additional coursework in computational neuroscience, machine learning, and deep learning. My master’s thesis, "Mechanisms of Opinion Change in Interacting Large Language Models: A Complex Systems Perspective", examined belief updating in interacting artificial agents. In a separate research project, I investigated the conditions under which subliminal learning occurs in neural networks.

The mechanistics of local-learning rules for curiosity in neural network models

My PhD investigates how local signals in neural networks can give rise to curiosity-like behaviour. I examine how neural systems act using their current internal model and how they process or seek information to improve it under biological and computational constraints. The goal is to uncover principles that support adaptive exploration and learning.

I am fascinated by how biological and artificial systems acquire, process, and use information, and by the broader question of what makes a system "intelligent". With my background in the physics of complex systems, I am especially interested in connecting microscopic interactions to macroscopic behaviour and in developing models that bridge these levels. In the context of curiosity, I aim to understand both the mechanisms that underlie it and how processes of self-organization may have given rise to them.

• Brockers, V. C., Ventzke, R. D., Neuhaus, V., Hidalgo-Ogalde, B., Priesemann, V. (2026). Learning Through Noise: Why Subliminal Learning Works and When It Fails.
• Brockers, V. C., Ehrlich, D. A., Priesemann, V. (2025). Disentangling Interaction and Bias Effects in Opinion Dynamics of Large Language Models.

From November 2023 to July 2026, I worked as a research assistant at the Max Planck Institute for Dynamics and Self-Organization. I developed pipelines for large-scale simulations, statistical inference, and model evaluation. I also taught undergraduate physics at the University of Göttingen. As a student tutor, I contributed to courses in mechanics, electromagnetism, mathematics, and mathematical methods in physics, as well as preparatory courses for incoming students.

• IC2S2, Norrköping, 2025: Plenary lightning talk and poster, Bayesian Modeling of Multi-Step Discussions Between LLM Agents: Disentangling Opinion Dynamics from Intrinsic Bias Effects.
• DPG Spring Meeting, Regensburg, 2025: Contributed talk, A Framework for Multi-Step Discussions of LLM-Based Agents.
• DPG Spring Meeting, Regensburg, 2025: Co-organizer of the focus session Large Language Models, Social Dynamics, and Assessment of Complex Systems.
• DPG Spring Meeting, Berlin, 2024: Poster, Opinion Dynamics of Interacting Large Language Models.