LLMs + Knowledge Graphs Improve ML Model Interpretability in Manufacturing

arXiv·medium signal

Bayer et al. present a method combining LLMs with domain-specific knowledge graphs to generate user-friendly explanations of ML model predictions in manufacturing settings. The approach bridges the gap between technical SHAP/LIME outputs and explanations production engineers can actually act on. Applicable to any domain where ML predictions need human-understandable justification beyond feature importance scores.