Atlas Kazemian

Distinct yet neighboring neural populations encode past, future, and surrounding speech context in the human temporal lobe

The paper shows that distinct neural populations in the human temporal lobe encode speech context from past, future, and surrounding input, with implications for speech comprehension models.

Convolutional architectures are cortex-aligned de novo

The paper shows that convolutional neural networks with minimal training can produce cortex-aligned visual representations through specific dimensionality manipulations, suggesting architectural constraints are key.