analiza aiqbench
This study rigorously disentangles genuine user-specific adaptation from mere improvements in population model strength within EEG foundation models for brain-computer interfaces. The findings demonstrate that while personal adapters consistently outperform both baseline population models and swapped parameters, the magnitude of this benefit is sensitive to training budget and unreliable when acquired via cheap few-label calibration methods. For practitioners deploying these models, the key takeaway is the necessity of evaluating personalization gains against robust population references rather than assuming inherent adaptability.
Separating personal from population gains when calibrating EEG foundation models for new users
arXiv:2609.34801v1 Announce Type: new Abstract: Foundation models are increasingly adapted to individual users, but an apparent personalization gain can simply reflect a stronger population model. This distinction matters for brain-computer interfaces, where every new user must be calibrated. We evaluated personal adaptation of three frozen EEG foundation models (CBraMod, REVE and LaBraM) in 235 held-out subjects from three motor-imagery datasets, comparing each subject's adapter with the population model and with adapters fitted to other subjects. Using all first-half session labels, personal adapters improved mean balanced accuracy over the population model by 1.5-5.4 percentage points and outperformed exchanged adapters by 2.3-7.3 points in all nine model-dataset combinations. The size of this benefit depended on population training: with four times the original budget, median gains remained positive (1.0-2.0 points) but were smaller for every model, and no population model reached a confirmed plateau. Acquiring the benefit cheaply was unreliable: few-label calibration was consistently non-negative on only one dataset, and in CBraMod neither unlabeled context nor meta-learned initialization outperformed matched controls. Personalization should therefore be evaluated against both a population reference and exchanged parameters, across population-training budgets.
Abstract:Foundation models are increasingly adapted to individual users, but an apparent personalization gain can simply reflect a stronger population model. This distinction matters for brain-computer interfaces, where every new user must be calibrated. We evaluated personal adaptation of three frozen EEG foundation models (CBraMod, REVE and LaBraM) in 235 held-out subjects from three motor-imagery datasets, comparing each subject's adapter with the population model and with adapters fitted to other subjects. Using all first-half session labels, personal adapters improved mean balanced accuracy over the population model by 1.5-5.4 percentage points and outperformed exchanged adapters by 2.3-7.3 points in all nine model-dataset combinations. The size of this benefit depended on population training: with four times the original budget, median gains remained positive (1.0-2.0 points) but were smaller for every model, and no population model reached a confirmed plateau. Acquiring the benefit cheaply was unreliable: few-label calibration was consistently non-negative on only one dataset, and in CBraMod neither unlabeled context nor meta-learned initialization outperformed matched controls. Personalization should therefore be evaluated against both a population reference and exchanged parameters, across population-training budgets.
59 pages, 19 figures (6 main, 13 supplementary), 33 tables (3 main, 30 supplementary). Includes supplementary information and source data. Code: this https URL
Machine Learning (cs.LG); Signal Processing (eess.SP); Neurons and Cognition (q-bio.NC)
arXiv:2609.34801 [cs.LG]
(or arXiv:2609.34801v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2609.34801 arXiv-issued DOI via DataCite (pending registration)
Text integral sub licența deschisă a publicației · 228 cuvinte
Subiecte conexe
[AINews] Meta Muse Wonder & Wonder Full: MSL’s Claude 5.5-tier Frontier Model
Latent Space · 13 oct. 2026
State of the Field: AI for Epistemics and Coordination
LessWrong · 2 oct. 2026
The J-Space Debate, Agent Swarms, and Pacing Frontier AI - Digital Minds Newsletter #4
LessWrong · 18 sept. 2026
A Red Line and Oversight Framework for Government AI Contracts
AI Alignment Forum · 18 iul. 2026