Researchers have found a way to improve the alignment between how artificial neural networks (ANNs) process information and how biological brains do. They used a new computational framework that adjusts the geometry of ANNs' internal representations during training. This led to better predictions in both directions: not just from ANNs to brain responses, but also from brain responses to ANN outputs. The improvements were modest, with a 55% increase in bidirectional predictivity. The study suggests that adjusting the geometry of neural networks can influence how well they align with biological brains. --- Why it matters: This research matters because it could lead to more accurate and efficient artificial intelligence models by better understanding how they relate to human cognition. Source: https://arxiv.org/abs/2608.18244