Linear Relational Decoding of Morphology in Language Models

AbstractA two-part affine approximation has been found to be a good approximation for transformer computations over certain subject-object relations. Adapting the Bigger Analogy Test Set, we show that the linear transformation W s , where s is a middle-layer representation of a subject token and W is derived from model derivatives, can accurately reproduce final object states for many relations. This linear technique achieves 90% faithfulness on morphological relations, with similar findings across languages and models. Our results suggest that some conceptual relationships in language models, such as morphology, are readily interpretable from latent space and are sparsely encoded by cross-layer linear transformations.