Shinji Nishimoto

This study uses encoding models to show that language and mathematics share some neural representations but also have distinct cortical organization.

This paper benchmarks the information content of fMRI, MEG, and EEG in visual tasks using a unified decoding framework, finding fMRI has the highest accuracy, MEG intermediate, and EEG improved with group-level aggregation.

The paper examines how second language proficiency affects brain-LLM alignment across neural layers, finding that higher proficiency correlates with stronger alignment in specific brain regions.

This study investigates how selective attention to different modalities modulates semantic feature decodability in brain activity using fMRI data from six participants.

This study uses fMRI to analyze cognitive task representations across 25 tasks, finding subject-specific and subject-invariant structures in cortical regions.

This study investigates whether symmetric semantic information is represented symmetrically in the human brain using fMRI data from movie-evoked responses.

This paper demonstrates that closed-loop brain-computer interfaces can retrieve intended images using a CLIP latent space, achieving over 80% accuracy, while failing with AlexNet.

This study uses large language models to predict human brain activity during movie watching, showing better accuracy for complex stories and identifying brain regions for multi-modal semantic representations.

This study improves image reconstruction from brain activity by incorporating semantic information, enabling reconstruction of both seen and imagined images.

The paper uses fMRI to show separate brain representations for language production and comprehension during natural dialogue, with production linked to motor systems and comprehension to auditory systems, both involving shared semantic and pragmatic networks.

This paper reconstructs high-resolution images from human brain activity using a latent diffusion model, showing high fidelity without additional training.

This study uses fMRI and artificial neural networks to show shared representations between ANN and BNN in mathematical operations, with deeper ANN layers improving decoding of novel operators.

This study uses fMRI and encoding models to show that mathematical problems are represented in the brain in a format-invariant way, regardless of being presented in symbolic or natural language form.

This paper introduces a framework using CNNs to decode individual mental information from predicted brain responses, achieving high accuracy without requiring additional fMRI measurements.

The paper quantifies how brain networks decouple from external stimuli during internal cognition using fMRI and a new coupling measure.

The paper investigates whether cognitive function representations are preserved across the cortex, cerebellum, and subcortex using fMRI data and decoding models.

The paper builds a model to predict human brain activity from video using motion-energy and TimeSformer features, finding that hybrid features predict whole brain activity best.

The paper compares semantic dissimilarities of words derived from brain models and human behavior, finding a significant correlation that supports the validity of word vector-based semantic representations in the brain.

This study examines modality invariance and attention selectivity in the human brain using fMRI, finding that perisylvian regions show both properties and that semantic features are more strongly associated with these phenomena than phonemic features.

The paper identifies brain regions associated with programming expertise by decoding source code categories from brain activity and correlating decoding accuracy with behavioral performance.

This study identifies 25 emotion dimensions in the human brain using fMRI and maps emotion categories on the cortical surface.

This study examines semantic representations in schizophrenia patients using fMRI and finds reduced intra- and inter-individual variability compared to healthy controls.

The paper uses fMRI and encoding models to reveal hierarchical organization of cognitive tasks and their cortical mapping.

The paper investigates whether topographic preservation of functional connectivity exists beyond the sensorimotor cortex, finding it also occurs in the association cortex.

The paper examines how music genres are represented in the human brain using fMRI and voxel-wise modeling, finding distinct cortical organizations in the STG explained by spectro-temporal modulation.