# Benchmarking Label-Revealed Online Updates for EEG BCI Decoding

Preprint · Bogdan Kozyrskiy, Artem Grachev, Abraham I. Camelo Guerrero

- Archive id: arxiv:2610.07420
- Published: 2026-10-05
- arXiv: https://arxiv.org/abs/2610.07420
- Citations: 0
- Source URL: https://arxiv.org/abs/2610.07420
- PDF: https://arxiv.org/pdf/2610.07420
- Provenance: arxiv

## Abstract

Electroencephalography (EEG) signals drift over time, which can cause static brain-computer interface (BCI) models to degrade in practice. We present a benchmark for online adaptation and compare two widely used pipeline families, Common Spatial Patterns (CSP) and Riemannian covariance-based methods, under time-ordered prequential (test-then-train) evaluation. We examine (i) which pipelines benefit most from label-revealed updates, (ii) whether controlled forgetting of older data improves robustness, and (iii) how a minimal-calibration cold start compares with starting from a pretrained model. Across four datasets (three motor-imagery datasets and one movement-decoding dataset), label-revealed online updates improve 13 of 14 model/dataset pairs on the two largest streams, with relative accuracy gains of up to about 18% over a frozen model. A Shapley-based data-valuation analysis over temporal blocks assigns the largest mean value to the most recent block in each of the three analyzed datasets, while older blocks retain positive value.

_Abstract reproduced under CC0-1.0 (arXiv descriptive metadata)._

Topics: [Brain-Computer Interfaces](https://space-kitty.com/topics/brain-computer-interfaces)

Cited in dossiers:
- [Brain-Computer Interfaces: State of the Field (October 2026)](https://space-kitty.com/dossiers/brain-computer-interfaces)

---
Space Kitty Archive · https://space-kitty.com · API: https://space-kitty.com/api/v1 · MCP: https://space-kitty.com/mcp · Premium: $0.20/request via x402 (Base Sepolia (testnet)). Metadata: arXiv (CC0), OpenAlex (CC0), Europe PMC (CC BY/CC0 works), ClinicalTrials.gov.