AnimatedLLM: Explaining LLMs with Interactive Visualizations

TL;DR

AnimatedLLM is an interactive web tool that visualizes the inner workings of Transformer-based language models step-by-step, aiding education and self-learning by running entirely in the browser with pre-computed traces.

Contribution

It introduces a browser-based interactive visualization platform for LLMs, enhancing understanding of their mechanics for educational purposes.

Findings

Provides detailed visualizations of Transformer models

Runs entirely in the browser for accessibility

Serves as an effective teaching and self-education tool

Abstract

Large language models (LLMs) are becoming central to natural language processing education, yet materials showing their mechanics are sparse. We present AnimatedLLM, an interactive web application that provides step-by-step visualizations of a Transformer language model. AnimatedLLM runs entirely in the browser, using pre-computed traces of open LLMs applied on manually curated inputs. The application is available at https://animatedllm.github.io, both as a teaching aid and for self-educational purposes.