Nora Belrose

Sparse Autoencoders Trained on the Same Data Learn Different Features

Evaluating SAE interpretability without generating explanations

Mitigating Emergent Misalignment with Data Attribution

Can We Partially Rewrite Transformers in Natural Language?

Automatically Interpreting Millions of Features in Large Language Models

Automatically Interpreting Millions of Features in Large Language Models

Balancing Label Quantity and Quality for Scalable Elicitation

Understanding Gradient Descent through the Training Jacobian