LLMs Unplugged: Teaching Resources for a ChatGPT World

Abstract

Large Language Models (LLMs) are everywhere, yet many learners lack a concrete mental model of how they generate text. This paper presents LLMs Unplugged, an unplugged set of activities that teaches the training-to-generation loop (and beyond) using hand-built ngram models and simple weighted sampling. Workshops based on these resources have been delivered to over 400 participants across secondary, tertiary, and executive-education contexts, and participants report that the activities demystify LLMs by reframing them as probabilistic "next word generation" at scale. All resources are freely available under a Creative Commons license at www. llmsunplugged.org, with a modular design that supports anything from a one hour crash course to a several-day intensive workshop.