kaballas/chess_llm_interpretability
Visualizing the internal board state of a GPT trained on chess PGN strings, and performing interventions on its internal board state and representation of player Elo.
Jupyter Notebook
No issues in this repository yet.