Sample-Efficient Behavior Cloning Using General Domain Knowledge

Feiyu Zhu, Jean Oh, Reid Simmons · 2025

Read Sample-Efficient Behavior Cloning Using General Domain Knowledge by Feiyu Zhu, Jean Oh, Reid Simmons (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-16546.txt.

Description

Feiyu Zhu, Jean Oh, and Reid Simmons show how plain-language know-how can help a robot learn a task from very few examples, in a 2025 paper. Copying an expert's moves, called behavior cloning, usually needs many demonstrations and fails on new situations. Experts can state general rules easily, but blending those rules with learned examples has been hard.

Their method asks a large language model to turn the expert's written rules into the skeleton of a control program, written as code. The few demonstrations then tune the numbers inside that skeleton. On lunar landing and car racing tasks, the system learned from as few as five demonstrations and stayed steady when noise was added to its actions. It beat a plain cloning model trained the same way.

Catalog record

CREATORFeiyu Zhu, Jean Oh, Reid Simmons
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE56 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-16546

Browse the full library · About the Tower of Babel