Ask HN: Why GPT should be reminded of the connection between some topics?

Sometimes, LLMs need reminders about the link between related topics (e.g., A and B), even though this connection exists in their training data. I suspect that while A is often mentioned alongside B, the reverse is less common. Are there practical techniques to reinforce these connections?