Martin Wattenberg

Into the Rabbit Hull: From Task-Relevant Concepts in DINO to Minkowski Geometry

Why Can't Transformers Learn Multiplication? Reverse-Engineering Implicit Chain-of-Thought Reveals Challenges of Learning Long-Range Dependencies

Beyond the Rosetta Stone: Unification Forces in Generalization Dynamics

Can Interpretation Predict Behavior on Unseen Data?

The Geometry of Self-Verification in a Task-Specific Reasoning Model

Shared Global and Local Geometry of Language Model Embeddings

Archetypal SAE: Adaptive and Stable Dictionary Learning for Concept Extraction in Large Vision Models

ICLR: In-Context Learning of Representations

Dialogue Action Tokens: Steering Language Models in Goal-Directed Dialogue with a Multi-Turn Planner

When Bad Data Leads to Good Models

What Does it Mean for a Neural Network to Learn a "World Model"?