As LLMs become larger, more capable, and more ubiquitous, the field of mechanistic interpretability  -- that is, understanding the inner wor...