Sparse Autoencoders Find Highly Interpretable Features in Language Models
Paper
•
2309.08600
•
Published
•
14
HaizeLabs
company
AI & ML interests
LLM Redteaming, security, robustness