
The authors propose a novel vocabulary-defined approach to analyze the semantics of language model latent space, which establishes a disentangled reference frame and enables effective model adaptation through semantic calibration.


coremsg

on-the-semantics-of-lm-latent-space-a-vocabulary-defined-approach-for-enhancing-language-model-performance-and-interpretability


On the Semantics of LM Latent Space: A Vocabulary-Defined Approach for Enhancing Language Model Performance and Interpretability
