Reproduce “Declarative Reactive Robot Behaviors from Imperative Language Models” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Vincent Berenz, Tasha Kim, Bernhard Schölkopf

PLAYFUL lets an LLM generate open-ended reactive robot behaviors by placing condition–action rules in a continuously evaluated script, while Python modules implement sensing and actions. With expert-authored guidelines, the agent placed more logic in the script than Python: across two Pepper behaviors, counts were 76/19, versus 40/41 without guidelines. These counts measure logic placement, not behavior quality; tests covered Pepper.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.