14 papers
Tensor Product Representation Probes Reveal Shared Structure Across Linear Directions
Andrew Lee, Fernanda Viégas, Martin Wattenberg
While researchers are finding concepts represented as linear directions in language models, a bag of linear directions fails to capture relational structure. To better understand t…
Decomposing Query-Key Feature Interactions Using Contrastive Covariances
Andrew Lee, Yonatan Belinkov, Fernanda Viégas +1
Despite the central role of attention heads in Transformers, we lack tools to understand why a model attends to a particular token. To address this, we study the query-key (QK) spa…
Why Can't Transformers Learn Multiplication? Reverse-Engineering Reveals Long-Range Dependency Pitfalls
Xiaoyan Bai, Itamar Pres, Yuntian Deng +5
Language models are increasingly capable, yet still fail at a seemingly simple task of multi-digit multiplication. In this work, we study why, by reverse-engineering a model that s…
Chronotome: Real-Time Topic Modeling for Streaming Embedding Spaces
Matte Lim, Catherine Yeh, Martin Wattenberg +2
Many real-world datasets -- from an artist's body of work to a person's social media history -- exhibit meaningful semantic changes over time that are difficult to capture with exi…
Story Ribbons: Reimagining Storyline Visualizations with Large Language Models
Catherine Yeh, Tara Menon, Robin Singh Arya +4
Analyzing literature involves tracking interactions between characters, locations, and themes. Visualization has the potential to facilitate the mapping and analysis of these compl…
What Does it Mean for a Neural Network to Learn a "World Model"?
Kenneth Li, Fernanda Viégas, Martin Wattenberg
We propose a set of precise criteria for saying a neural net learns and uses a "world model." The goal is to give an operational meaning to terms that are often used informally, in…