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.
Tim Fernholz
6 views
