MINT: Minimal Information Neuro-Symbolic Tree for Objective-Driven Knowledge-Gap Reasoning and Active Elicitation

Researchers introduce MINT (Minimal Information Neuro-Symbolic Tree), a framework enabling AI agents to actively request human input during collaborative planning. MINT combines symbolic reasoning with neural networks to identify knowledge gaps—unknown objects, goals, or constraints—and optimally queries humans to fill them. By analyzing planning consequences through a symbolic tree and consulting neural policies, MINT generates targeted questions that improve planning performance with minima...