NeuroCLIP: Brain-Inspired Prompt Tuning for EEG-to-Image Multimodal Contrastive Learning

TL;DR

NeuroCLIP introduces a brain-inspired prompt tuning framework for EEG-to-image contrastive learning, enhancing neural-visual alignment through adaptive prompts and a neuroscientific-inspired loss, achieving significant improvements in zero-shot image retrieval.

Contribution

The paper presents the first integration of visual prompt tokens into EEG-image alignment and a dual-stream adaptive prompt design inspired by neuroscience principles.

Findings

Achieved 63.2% Top-1 accuracy in zero-shot image retrieval on THINGS-EEG2 dataset.

Surpassed previous methods by +12.3% in accuracy.

Demonstrated strong generalization across subjects (+4.6% Top-1).

Abstract

Recent advances in brain-inspired artificial intelligence have sought to align neural signals with visual semantics using multimodal models such as CLIP. However, existing methods often treat CLIP as a static feature extractor, overlooking its adaptability to neural representations and the inherent physiological-symbolic gap in EEG-image alignment. To address these challenges, we present NeuroCLIP, a prompt tuning framework tailored for EEG-to-image contrastive learning. Our approach introduces three core innovations: (1) We design a dual-stream visual embedding pipeline that combines dynamic filtering and token-level fusion to generate instance-level adaptive prompts, which guide the adjustment of patch embedding tokens based on image content, thereby enabling fine-grained modulation of visual representations under neural constraints; (2) We are the first to introduce visual prompt tokens into EEG-image alignment, acting as global, modality-level prompts that work in conjunction with instance-level adjustments. These visual prompt tokens are inserted into the Transformer architecture to facilitate neural-aware adaptation and parameter optimization at a global level; (3) Inspired by neuroscientific principles of human visual encoding, we propose a refined contrastive loss that better model the semantic ambiguity and cross-modal noise present in EEG signals. On the THINGS-EEG2 dataset, NeuroCLIP achieves a Top-1 accuracy of 63.2% in zero-shot image retrieval, surpassing the previous best method by +12.3%, and demonstrates strong generalization under inter-subject conditions (+4.6% Top-1), highlighting the potential of physiology-aware prompt tuning for bridging brain signals and visual semantics.