About this document

Visual Analytics for Transformer Models by Jorge Casajus is a document available to read on EtoBox.

This paper proposes a visual analytics framework to help analyze and interpret transformer-based generative models. The framework visualizes hidden state projections, attention head importance, and instance-level input attribution and attention weights. It aims to support tasks like summarization, translation, and question answering by providing more insight into these complex models. The framework is one of the first dedicated to visualizing encoder-decoder and decoder-only generative models. It demonstrat

Author

Jorge Casajus

Language

EN