Scholarlink

We present a novel neuro-symbolic framework that enhances BioBERT-based biomedical question answering (QA) by integrating real-time EEG signals and symbolic algebra into the inference process of large language models (LLMs). Leveraging the PubMedQA dataset and pretrained biomedical transformers such as BioBERT, our system models the user’s cognitive responses using EEG data from PhysioNet and computes inter-channel correlation matrices to construct symbolic representations of brain activity. These symbolic constructs guide the model’s attention by injecting token-level biases, thus aligning inference with cognitively salient biomedical terms. This transparent augmentation allows for interpretable and brain-aligned answers, advancing the frontier of explainable AI (XAI) in healthcare. Our experiments demonstrate that signal-aware attention modulation improves answer relevance and interpretability, paving the way for adaptive human-AI interaction in neurocognitive and clinical domains. The implementation accompanying this work is publicly available at https://github.com/satyamcser/signal-algebra-llms-brain-qa .

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Neuro-Symbolic Attention Modulation in Biomedical QA via Signal Algebra-Augmented LLMs

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ha Yeon Sohn,
  • Satyam Mishra,
  • Vishwanath Bijalwan

摘要

We present a novel neuro-symbolic framework that enhances BioBERT-based biomedical question answering (QA) by integrating real-time EEG signals and symbolic algebra into the inference process of large language models (LLMs). Leveraging the PubMedQA dataset and pretrained biomedical transformers such as BioBERT, our system models the user’s cognitive responses using EEG data from PhysioNet and computes inter-channel correlation matrices to construct symbolic representations of brain activity. These symbolic constructs guide the model’s attention by injecting token-level biases, thus aligning inference with cognitively salient biomedical terms. This transparent augmentation allows for interpretable and brain-aligned answers, advancing the frontier of explainable AI (XAI) in healthcare. Our experiments demonstrate that signal-aware attention modulation improves answer relevance and interpretability, paving the way for adaptive human-AI interaction in neurocognitive and clinical domains. The implementation accompanying this work is publicly available at https://github.com/satyamcser/signal-algebra-llms-brain-qa .