Solving the Mechanistic Interpretability challenges: EIS VII Challenge 2 — AI Alignment Forum