openai / automated-interpretability

940 stars 110 forks source link

Add a note about a bug in how GPT-2 activations were calculated #22

Closed stevenbills closed 1 year ago