5) Are there are other formal or rule-based domains where including its data in LLM pretraining can help?

→ View original post on X — @askalphaxiv