Sparse Autoencoders Find Highly Interpretable Features in Language Models Paper • 2309.08600 • Published Sep 15, 2023 • 15
