Guardado en:
Detalles Bibliográficos
Autores principales: Choi, Hanseul, Park, Jinyeong, Jin, Seongwon, Park, Sungho, Kim, Jibum
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:https://arxiv.org/abs/2603.13272
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866910052452925440
author Choi, Hanseul
Park, Jinyeong
Jin, Seongwon
Park, Sungho
Kim, Jibum
author_facet Choi, Hanseul
Park, Jinyeong
Jin, Seongwon
Park, Sungho
Kim, Jibum
contents Electroencephalography (EEG) foundation models have shown promise for learning generalizable representations, yet they remain sensitive to channel heterogeneity, such as changes in channel composition or ordering. We propose channel-aware multimodal EEG-text alignment contrastive language-image pretraining (CAMEL-CLIP), a contrastive EEG-text multimodal foundation model designed to be robust to heterogeneous channel configurations and widely applicable to diverse downstream tasks. CAMEL-CLIP introduces three key components: (1) channel attribute-based positional encoding, which identifies channels through semantic information; (2) dynamic channel projection, which generates variable-length embeddings by independently projecting each channel without feature compression; and (3) dual-level contrastive learning, which jointly performs channel-level and sample-level contrastive learning to capture both channel-specific and global signal characteristics. Experimental results demonstrate that CAMEL-CLIP achieves state-of-the-art performance under linear-probing and outperforms existing foundation models that rely on full-finetuning.
format Preprint
id arxiv_https___arxiv_org_abs_2603_13272
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle CAMEL-CLIP: Channel-aware Multimodal Electroencephalography-text Alignment for Generalizable Brain Foundation Models
Choi, Hanseul
Park, Jinyeong
Jin, Seongwon
Park, Sungho
Kim, Jibum
Machine Learning
Artificial Intelligence
Electroencephalography (EEG) foundation models have shown promise for learning generalizable representations, yet they remain sensitive to channel heterogeneity, such as changes in channel composition or ordering. We propose channel-aware multimodal EEG-text alignment contrastive language-image pretraining (CAMEL-CLIP), a contrastive EEG-text multimodal foundation model designed to be robust to heterogeneous channel configurations and widely applicable to diverse downstream tasks. CAMEL-CLIP introduces three key components: (1) channel attribute-based positional encoding, which identifies channels through semantic information; (2) dynamic channel projection, which generates variable-length embeddings by independently projecting each channel without feature compression; and (3) dual-level contrastive learning, which jointly performs channel-level and sample-level contrastive learning to capture both channel-specific and global signal characteristics. Experimental results demonstrate that CAMEL-CLIP achieves state-of-the-art performance under linear-probing and outperforms existing foundation models that rely on full-finetuning.
title CAMEL-CLIP: Channel-aware Multimodal Electroencephalography-text Alignment for Generalizable Brain Foundation Models
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2603.13272