Scholarlink

<p>This short communication presents the explainable neuro-symbolic framework (ENSF), a novel architecture integrating differentiable logical inference with attention-guided explanation generation. Unlike existing neuro-symbolic approaches that sacrifice transparency for accuracy, ENSF’s primary innovation is a unified pipeline enabling real-time symbolic rule adaptation while maintaining interpretable outputs. Evaluated across autonomous driving, medical diagnosis, and industrial robotics (5 independent runs each), ENSF achieves a 15% improvement in adversarial robustness and 20% reduction in out-of-distribution errors versus state-of-the-art baselines (ResNet-50, Logic Tensor Networks), with 85% explanation fidelity validated by 12 domain experts. ENSF addresses critical requirements for AI deployment in safety–critical, regulated environments.</p>

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

Explainable Neuro-Symbolic Architectures for Robust and Adaptive Intelligence in Complex Real-World Scenarios

  • 期刊论文

  • 发表日期：

    2026年5月17日

查看全文
  • Vishal Tiwari,
  • Vinay Keswani

摘要

This short communication presents the explainable neuro-symbolic framework (ENSF), a novel architecture integrating differentiable logical inference with attention-guided explanation generation. Unlike existing neuro-symbolic approaches that sacrifice transparency for accuracy, ENSF’s primary innovation is a unified pipeline enabling real-time symbolic rule adaptation while maintaining interpretable outputs. Evaluated across autonomous driving, medical diagnosis, and industrial robotics (5 independent runs each), ENSF achieves a 15% improvement in adversarial robustness and 20% reduction in out-of-distribution errors versus state-of-the-art baselines (ResNet-50, Logic Tensor Networks), with 85% explanation fidelity validated by 12 domain experts. ENSF addresses critical requirements for AI deployment in safety–critical, regulated environments.