This research introduces a framework that leverages Large Language Models to help robots accurately rearrange objects based on natural language instructions and past experiences.

• Employs LLMs to interpret language commands and determine precise object placement
• Enables robots to learn from previous arrangements rather than relying solely on pre-collected datasets
• Significantly improves rearrangement accuracy and reduces task complexity
• Creates more adaptable robot systems capable of handling varied instructions beyond training scenarios

For engineering applications, this approach represents a breakthrough in human-robot collaboration, allowing industrial robots to understand natural language commands and execute complex rearrangement tasks with greater precision and flexibility.

Learn from the Past: Language-conditioned Object Rearrangement with Large Language Models