Lucas Benjamin

The paper shows humans make more accurate decisions by compressing sensory evidence into decision categories, and RNNs trained on the same task exhibit similar resilience to noisy computations.

This study shows human perceptual decisions use abstract neural representations to flexibly combine sensory features across contexts, supported by MEG data and recurrent neural networks.

The paper shows that humans encode sequences with order-based structure as mental programs, improving recall accuracy and predicting motor production patterns.

This paper introduces a unifying framework for statistical learning across temporal scales, showing a single mechanism can capture both local and higher-order patterns in data.

The paper investigates how sensory integration, temporal prediction, and rule discovery interact in humans using rhythmic stimulus sequences and finds they operate interdependently.

This study investigates the neural mechanisms of geometric shape perception in humans using fMRI and MEG, finding distinct brain activation patterns and suggesting a symbolic mode of visual processing.

The paper examines neural speech processing differences in infants at high and low likelihood for autism using EEG, finding reduced syllable tracking in high likelihood infants linked to verbal outcomes.

The paper investigates whether neonates can detect statistical regularities in speech, showing they process both phonetic and voice dimensions and respond specifically to phonetic regularities.

The paper examines how environmental and biological factors influence functional connectivity in the superior temporal sulcus (STS) of neonates, revealing sex and prematurity effects on brain network development.

The paper investigates how humans process auditory sequences with network structures using MEG, finding rapid encoding of network structure and a long-horizon associative learning mechanism.

This study investigates whether statistical learning occurs in patients with impaired consciousness by analyzing their ability to extract regularities from an artificial language using EEG frequency tagging.

The paper shows humans use compressed network models to represent auditory sequences, pruning and completing relationships to efficiently capture both local and high-order structures.

This study shows that tracking transitional probabilities and segmenting auditory sequences are separate processes in adults and neonates, with prosodic cues aiding segmentation and learning increasing neural similarity across infants.

The paper investigates how neonates process and retain an artificial language, finding a two-step learning strategy involving neural entrainment and memory of syllable sequences.

APICE is an automated EEG preprocessing pipeline for infants that adapts to different ages and experimental conditions, validated on two datasets.

The paper shows that microsaccade inhibition within 300ms of sound onset is a reliable indicator of auditory salience, validated through crowd-sourced ratings and neural measurements.