MIND: Instruction-Conditioned Incremental Grounding for Robots

MIND is a framework that enables robots to maintain target-room decisions without re-evaluating every room after each event. It uses a language model to compile instructions into dependencies, and a controller to intersect these with event footprints, refreshing judgments and updating reachability and selection. MIND reduces mean semantic evaluations from 30.6 to 11.2 in simulated environments while preserving all post-event decisions.