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.

from TechCrunch https://ift.tt/V57mfdF