openai / automated-interpretability

977 stars 116 forks source link