/sae_vis

Create feature-centric and prompt-centric visualizations for sparse autoencoders (like those from Anthropic's published research).

Primary LanguageHTMLMIT LicenseMIT

No issues in this repository yet.