Teach a Programming Assistant - SQL v2 - Helpful, Honest, Harmless
This project provides a set of example databases in SQLite and tasks the human with conversing with an AI assistant about analysis of one of those databases. Each conversation turn involves two responses from the assistant which the human then specifies which one they prefer and for what reason (Helpfulness, Harmlessness, Honesty). This iteration focuses on continued conversations, up to three turns, and asks the human to find methods that produce negative responses.