PISCES. πͺPISCES - Precise In-Parameter Suppression for Concept EraSure in Large Language Models
13Feature-Descriptions. Enhancing Automated Interpretability with Output-Centric Feature Descriptions
12mixing-mechs. Official code for "Mixing Mechanisms: How Language Models Retrieve Bound Entities In-Context"
9BonaFide. A benchmark for evaluating faithfulness metrics with ground-truth.
8mechinterp. Library for easily using interpretability techniques in transformer models
6yoavgur.github.io. HTML
1AutoBib. Plugin for copying published BibTeX to clipboard
1