This research introduces Introspective Planning - a novel approach that improves robot safety by aligning language model uncertainty with task ambiguity.

• Addresses the critical issue of LLM hallucination that can lead to robots executing unsafe actions
• Calibrates language model confidence to reflect genuine ambiguity in instructions
• Establishes a new benchmark for safe mobile manipulation
• Demonstrates significant improvements in both task compliance and safety

For security professionals, this research offers a promising framework to reduce risks in autonomous robotic systems where incorrect decision-making could lead to physical harm or security breaches.

Introspective Planning: Aligning Robots' Uncertainty with Inherent Task Ambiguity