LibriBrain100: One Hundred Hours of Broad and Deep MEG Data for Neural Speech Decoding at Scale
- 1PNPL , Department of Engineering Science , University of Oxford , UK
- 2FMRIB , Oxford Centre for Integrative Neuroimaging , University of Oxford , UK
- 3OHBA , Oxford Centre for Integrative Neuroimaging , University of Oxford , UK
Abstract
We introduce LibriBrain100, a large-scale MEG dataset for speech decoding designed from the ground up for reproducible, standardised evaluation. LibriBrain100 more than doubles the size of the original LibriBrain release, resulting in over 100 hours of high-quality MEG acquired while subjects listened to naturalistic continuous speech. With ∼80 hours from a single subject, LibriBrain100 sets a new record for deep, within-subject neural data (8× more than the next comparable dataset and roughly 80× more than other datasets). To demonstrate the payoff of this depth-first design, we evaluate on a word-classification benchmark---an increasingly well-established stepping stone towards the open challenge of noninvasive brain-to-text decoding. Using an existing decoding model, we achieve state-of-the-art performance---validating both the quality of the recordings and the value of within-subject data at scale. Because collecting 80 hours of data per user is impractical for real-world applications, we also collected ∼40 minutes of additional data from each of 32 subjects. Using the same word-classification benchmark, we demonstrate the value of broad multi-subject data: supervised finetuning of a pre-trained model can substantially compensate for limited per-subject data. We provide standard train, validation, and test splits, all reproducible through an open-sourced Python library that supports easy downloading, optional preprocessing, and data loading for common deep learning frameworks. In addition, the dataset and evaluation infrastructure are being released alongside an open machine-learning competition with a public leaderboard for standardised benchmarking. Ultimately, our hope is that LibriBrain100 will accelerate progress towards practical non-invasive brain-computer interfaces, capable of restoring communication to people living with severe paralysis.
Related
- Speech Restoration & Mental Surveillance as Conflicting Ethical Concerns in Neural Speech-Decoding Research.
- Shared latent representations of speech production for cross-patient speech decoding
- Simultaneous speech and gesture decoding for multimodal communication in paralysis
- Two Kinds of Mental Privacy: What Speech Decoding Does and Does Not Threaten.
- Simulated transcranial ultrasound of human language cortex: verifying heard speech needs far less data and signal than decoding it
BibTeX
@misc{mantegna2026libribrain100,
title = {{LibriBrain100: One Hundred Hours of Broad and Deep MEG Data for Neural Speech Decoding at Scale}},
author = {Francesco Mantegna and Dulhan Jayalath and Gereon Elvers and Tasha Kim and Benjamin Ballyk and Alex Fung and SungJun Cho and Teyun Kwon and Luisa Kurth and Miran Özdogan and Gilad Landau and Pratik Somaiya and Natalie Voets and Mark W. Woolrich and Ōiwi Parker Jones},
howpublished = {arXiv (Cornell University)},
year = {2026},
month = aug,
doi = {10.48550/arxiv.2608.25204},
url = {https://arxiv.org/abs/2608.25204},
note = {CC BY},
}
RIS
TY - UNPB
TI - LibriBrain100: One Hundred Hours of Broad and Deep MEG Data for Neural Speech Decoding at Scale
AU - Francesco Mantegna
AU - Dulhan Jayalath
AU - Gereon Elvers
AU - Tasha Kim
AU - Benjamin Ballyk
AU - Alex Fung
AU - SungJun Cho
AU - Teyun Kwon
AU - Luisa Kurth
AU - Miran Özdogan
AU - Gilad Landau
AU - Pratik Somaiya
AU - Natalie Voets
AU - Mark W. Woolrich
AU - Ōiwi Parker Jones
PY - 2026
DA - 2026/08/25
PB - arXiv (Cornell University)
DO - 10.48550/arxiv.2608.25204
UR - https://arxiv.org/abs/2608.25204
AB - We introduce LibriBrain100, a large-scale MEG dataset for speech decoding designed from the ground up for reproducible, standardised evaluation. LibriBrain100 more than doubles the size of the original LibriBrain release, resulting in over 100 hours of high-quality MEG acquired while subjects listened to naturalistic continuous speech. With ∼80 hours from a single subject, LibriBrain100 sets a new record for deep, within-subject neural data (8× more than the next comparable dataset and roughly 80× more than other datasets). To demonstrate the payoff of this depth-first design, we evaluate on a word-classification benchmark---an increasingly well-established stepping stone towards the open challenge of noninvasive brain-to-text decoding. Using an existing decoding model, we achieve state-of-the-art performance---validating both the quality of the recordings and the value of within-subject data at scale. Because collecting 80 hours of data per user is impractical for real-world applications, we also collected ∼40 minutes of additional data from each of 32 subjects. Using the same word-classification benchmark, we demonstrate the value of broad multi-subject data: supervised finetuning of a pre-trained model can substantially compensate for limited per-subject data. We provide standard train, validation, and test splits, all reproducible through an open-sourced Python library that supports easy downloading, optional preprocessing, and data loading for common deep learning frameworks. In addition, the dataset and evaluation infrastructure are being released alongside an open machine-learning competition with a public leaderboard for standardised benchmarking. Ultimately, our hope is that LibriBrain100 will accelerate progress towards practical non-invasive brain-computer interfaces, capable of restoring communication to people living with severe paralysis.
SN - 2331-8422
ER -