Neuroscore: 17,175 Texts Scored by Predicted Brain Engagement

NeuroScore is a dataset of 17,175 text items scored by predicted human brain activation using Meta's TRIBE v2 brain encoding model. The model predicts fMRI responses based on training data from over 700 human subjects, providing activation scores across 6 cortical regions, a composite engagement score, and an emotion profile. The dataset was created by author tushar710 and last updated on Hugging Face in April 2026.

Use Cases

• Benchmarking text generation models based on predicted neural engagement scores.
• Training classifiers for emotional content based on the provided emotion profile classification.
• Analyzing text features that correlate with activation in specific cortical regions of interest (ROIs).
• Developing content recommendation systems based on predicted cognitive engagement.

Strengths

• Contains 17,175 text items, providing a substantial corpus for analysis.
• Scores derived from Meta's TRIBE v2 model trained on fMRI data from 700+ human subjects.
• Provides multiple predicted outputs per text: activation across 6 ROIs, a composite score, and an emotion profile.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is known, but data format, sample, and license details are unknown.
• Data may reflect bias inherent to the training data and model architecture of TRIBE v2.

Provenance

Source

Hugging Face dataset by author tushar710, using Meta's TRIBE v2 model.

Collection Method

Texts processed through the TRIBE v2 brain encoding model to predict fMRI-based responses.

Time Range

null

Freshness

Last updated 2026-04-11 15:50:35; freshness should be verified.

Geography

null