Gemma Scope: helping the safety community shed light on the inner workings of language models

Announcing a comprehensive, open suite of sparse autoencoders for language model interpretability.

Source: Read More