边际正则化结构化语义对齐用于脑语言对应研究

AI 摘要

本文提出 MD-SigLIP 框架，通过边际正则化的结构化语义对齐，将脑磁图信号与文本嵌入映射到共享语义空间，实现基于检索的脑语言解码。该方法在 D-SigLIP 基础上引入列表式边际正则项，建模多正例语义簇和排序约束，在两个 MEG 语言数据集上取得最优检索性能，且不依赖生成式语言模型。

正文节选

Margin-Regularized Structured Semantic Alignment for Brain-Language Correspondence Abstract With the rapid advancement of large language models, brain-language decoding has achieved remarkable progress. However, it remains unclear whether decoded content genuinely reflects neural representations or is largely reconstructed by the language model itself. This ambiguity limits interpretability and hinders the investigation of intrinsic brain-language correspondence. To address this challenge, we pr

发布时间：2026-08-19 12:00

抓取时间：2026-08-19 12:08

来源机构：arXiv