Relational Inference and knowledge composition via neuronal geometric representations

This workshop aims to explore how the brain constructs internal maps of implicit relationships between stimuli, and how task structure shapes the geometry of these representations. At the intersection of cognitive and computational neuroscience, we will discuss key questions about generalisation: What principles allow the brain to infer novel relations from prior knowledge? How are spatial representations intertwined with higher cognition? And can geometry serve as a fundamental coding paradigm for structured information? To address these themes, we bring together researchers approaching the problem from complementary perspectives: computational models of relational generalisation, experimental studies on inferential reasoning, and analyses of the geometric properties of neural representations. Through a series of talks and two panel discussions, we aim to foster dialogue between these communities, identify unifying principles, and outline open challenges that can shape future investigations into the neural basis of abstract, structured cognition.