Popular repositories Loading

1. SpellGPT SpellGPT Public

   An experimental tool to explore GPT-3's "miraculous" ability not only to spell its own token strings (it being a "character blind" model) but also to use spelling as a means to produce novel output…

   Python 11 1

2. GPT_definition_trees GPT_definition_trees Public

   code for generating "definition trees" for (not-necessarily-token) GPT-J embeddings

   Python 7 2

3. SAE_Feature_Interpretability_Tool SAE_Feature_Interpretability_Tool Public

   A tool to assist in the interpretation of learned features in sparse autoencoders (in particular the four SAE's trained by Joseph Bloom on Gemma-2B).

   Jupyter Notebook 6

4. alphabetical_probe alphabetical_probe Public

   Experimental code which trains 26 linear probes to detect the presence of alphabetic letters in GPT-J token strings, given their embeddings. Exploring the resulting vector arithmetic and its impact…

   Jupyter Notebook 2 2

5. Leilan-dataset Leilan-dataset Public

   Dataset containing 600 interview transcripts involving a GPT-3 simulacrum built around the ' Leilan' glitch token.

   Python 2

6. stampy-chat stampy-chat Public

   Forked from StampyAI/stampy-chat

   Conversational chatbot to answer questions about AI Safety & Alignment based on information retrieved from the Alignment Research Dataset

   Python