MTEEG: Multi-Task EEG Analysis with Low-Rank Adaptation

A new framework called MTEEG enables simultaneous adaptation of a pre-trained EEG model to multiple tasks using task-specific low-rank adaptation (LoRA) modules. This addresses the challenge of EEG signal heterogeneity across subjects, devices, and experimental setups, which causes conflicts during joint optimization. By disentangling the parameter space, MTEEG reduces computational and spatial costs compared to maintaining separate models for each task. The approach builds on recent self-supervised pre-training methods for EEG, which previously required full fine-tuning per task. The study is published on arXiv under ID 2604.25131.