Scaling subjects in cross-modal alignment: video decoding with EEG foundation model
Abstract
Naturalistic visual decoding from EEG has long been constrained by cohort size: existing scaling literature caps out at a few dozen subjects, leading prior work to conclude that scaling subject cohorts yields minimal performance gains. Training a cross-modal EEG--video contrastive encoder on a cohort larger by more than an order of magnitude, we find the axis is productive but has an onset. Below S≈50 --- the entirety of the range prior work occupies --- no model improves meaningfully over an untrained encoder; above it, decoding rises log-linearly in subject count with no saturation at the top of our ladder, on a fitted stimulus-feature probe and on fit-free movie-moment retrieval alike, and the gain transfers to a second, unseen film. How far the axis carries then depends on initialisation more than on capacity: an encoder initialised from an EEG foundation model scales ∼1.6× faster per doubling of the cohort than randomly initialised encoders at two depths, is the only one still converting subjects into retrieval accuracy at the top of the ladder, and converges on a fraction of the alignment compute. Subject scaling on shared naturalistic stimuli is thus an effective and currently unsaturated frontier for scaling EEG foundation models.
Related
- Limitations of Genomic Foundation Models for Decoding Regulatory Mechanisms in ALS
- Decoding Alignment without Encoding Alignment: A critique of similarity analysis in neuroscience
- FoME: A foundation model for EEG using adaptive temporal-lateral attention scaling
- Scaling recipes for single-cell RNA sequencing foundation models: when do scaling laws hold?
- Towards Trustworthy Biological Alignment in TabPFN-Probed Pathology Foundation Models
BibTeX
@misc{truong2026scaling,
title = {{Scaling subjects in cross-modal alignment: video decoding with EEG foundation model}},
author = {Dung Truong and Kuntal Kokate and Arnaud Delorme},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.09287},
eprint = {2610.09287},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.09287},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Scaling subjects in cross-modal alignment: video decoding with EEG foundation model
AU - Dung Truong
AU - Kuntal Kokate
AU - Arnaud Delorme
PY - 2026
DA - 2026/10/07
PB - arXiv
DO - 10.48550/arxiv.2610.09287
UR - https://arxiv.org/abs/2610.09287
AB - Naturalistic visual decoding from EEG has long been constrained by cohort size: existing scaling literature caps out at a few dozen subjects, leading prior work to conclude that scaling subject cohorts yields minimal performance gains. Training a cross-modal EEG--video contrastive encoder on a cohort larger by more than an order of magnitude, we find the axis is productive but has an onset. Below S≈50 --- the entirety of the range prior work occupies --- no model improves meaningfully over an untrained encoder; above it, decoding rises log-linearly in subject count with no saturation at the top of our ladder, on a fitted stimulus-feature probe and on fit-free movie-moment retrieval alike, and the gain transfers to a second, unseen film. How far the axis carries then depends on initialisation more than on capacity: an encoder initialised from an EEG foundation model scales ∼1.6× faster per doubling of the cohort than randomly initialised encoders at two depths, is the only one still converting subjects into retrieval accuracy at the top of the ladder, and converges on a fraction of the alignment compute. Subject scaling on shared naturalistic stimuli is thus an effective and currently unsaturated frontier for scaling EEG foundation models.
ER -