The Gradient: Perspectives on AI cover image

Jacob Andreas: Language, Grounding, and World Models

The Gradient: Perspectives on AI

00:00

Exploring Model Editing and Task Vector Dynamics

This chapter delves into novel methods for model editing, focusing on direct interventions with representations and parameters. It covers the use of task vectors and the implications of weight subtraction between models, while addressing uncertainties in parameter space that affect model truthfulness and fact verification.

Transcript
Play full episode

The AI-powered Podcast Player

Save insights by tapping your headphones, chat with episodes, discover the best highlights - and more!
App store bannerPlay store banner
Get the app