A cognition grounded interpretability method

We re-evaluate whether vision-and-language models exhibit the human-like bouba-kiki effects, using two methods modelled after human experiments. Compared to humans, VLMs fall short in aligning cross-modal associations with human intuitions.

We propose methods for studying behavior and neural dynamics in deep RL agents based on neuroscience methods and demonstrate them in a naturalistic simulated foraging environment

LibriBrain is the largest non-invasive MEG dataset (over 50 hours) recorded from a single subject listening to naturalistic speech, designed to advance scalable and reproducible machine learning methods for speech decoding from brain activity.