This research introduces a simple yet powerful framework that enables robots to plan complex tasks interactively using language models, adapting to new goals even during execution.

• Eliminates need for heavy prompt engineering or domain-specific models
• Implements function calling for effective robot-LLM integration
• Demonstrates real-world success in varied domains including gastronomy tasks
• Achieves generalization to new goals without additional training

For engineering teams, this approach offers a breakthrough in developing adaptable robotic systems that can understand natural language instructions and modify plans in real-time, potentially accelerating automation across various industries.

Interactive Task Planning with Language Models