MINT: Minimal Information Neuro-Symbolic Tree for Objective-Driven Knowledge-Gap Reasoning and Active Elicitation
Researchers have developed a novel neuro-symbolic approach called MINT (Minimal Information Neuro-Symbolic Tree) to address knowledge gaps in human-AI collaboration for planning tasks. MINT constructs a symbolic tree to estimate planning uncertainties caused by missing information and uses self-play to optimize AI elicitation strategies. The system leverages large language models to refine queries for human input, aiming to improve planning performance. AI
IMPACT Introduces a new method for AI agents to actively elicit information from humans, potentially improving collaborative planning in complex environments.