Do Language Models Encode Semantic Relations? Probing and Sparse Feature Analysis

Andor Diera, Ansgar Scherp

Understanding whether large language models (LLMs) capture structured meaning requires examining how they represent concept relationships. In this work, we study three models of in…