TidyBot: a tidying robot that learns your preferences via LLM
In one sentence Stanford presents TidyBot, a robotic system that uses LLMs to personalize household tidying behavior from a few user examples. It achieves 91.2% task completion, demonstrating the feasibility of LLM-driven personalization in manipulation.
Every home is different. Some people put cups in the right cabinet, others in the left. A generic tidying robot cannot know this. Stanford's TidyBot solves this problem cleverly: instead of programming fixed rules, it asks an LLM to infer user preferences from a few examples.
Here is how it works: you show the robot three or four examples of where you want things placed ("cups go in the cabinet", "kids' toys in the basket"), and the language model automatically generates behavioral rules for everything else. If you have a new cup it has never seen, the robot figures out where to put it based on the preferences it has already learned.
The results are impressive: 91.2% task completion. For a household robot learning from very few examples, this is excellent performance.
The breakthrough is not purely technical: TidyBot demonstrates that LLMs are not just for answering questions — they can become the "preference brain" of a physical robot. It is one of the first systems to combine natural language understanding and robotic manipulation in a genuinely domestic context.
Companies
Stanford
Tools
—
Tags
Sources