Guide Labs debuts a new kind of interpretable LLM
The company open-sourced an 8 billion parameter LLM, Steerling-8B, trained with a new architecture designed to make its actions easily interpretable.
steminsights.org
0