Create feature-centric and prompt-centric visualizations for sparse autoencoders (like those from Anthropic's published research).
By chatting or signing in you agree to the Terms and chat-message logging (revocable in History).