description evaluation

faithfulness metric (1) large language model (1) automated interpretability (1) feature description (1) sparse autoencoder (1) mechanistic interpretability (1) language model (1) latent representation (1) interpretability evaluation (1) feature alignment (1)