Linear Structure of High-Level Concepts in Text-Controlled Generative Models

Victor Veitch, University of Chicago

11.01.2024

5:15 pm - 6:45 pm

LMU Munich, Institute of AI in Management via zoom

Text controlled generative models (such as large language models or text-to-image diffusion models) operate by embedding natural language into a vector representation, then using this representation to sample from the model’s output space. This talk concerns how high-level semantics are encoded in the algebraic structure of representations.

#lecture

Subscribe to RSS Events feed

Colloquium • 06.05.2026 • LMU Munich, Department of Statistics and via zoom

Data Thinning and Beyond

06.05.26, 4:15-5:45 pm: Daniela Witten from the University of Washington

Lecture • 12.06.2026 • LMU Munich, CAS, Seestr. 13, Munich

Analyzing Feature Interactions Through Local Effects in Machine Learning Models

As part of the CAS Research Focus, Giuseppe Casalicchio talks about interpretable machine learning that develops methods.

Linear Structure of High-Level Concepts in Text-Controlled Generative Models

Victor Veitch, University of Chicago

Related