English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Language instructions and demonstrations are two natural ways for users toteach robots personalized tasks. Recent progress in Large Language Models(LLMs) has shown impressive performance in translating language instructionsinto code for robotic tasks. However, translating demonstrations into task codecontinues to be a challenge due to the length and complexity of bothdemonstrations and code, making learning a direct mapping intractable. Thispaper presents Demo2Code, a novel framework that generates robot task code fromdemonstrations via an extended chain-of-thought and defines a common latentspecification to connect the two. Our framework employs a robust two-stageprocess: (1) a recursive summarization technique that condenses demonstrationsinto concise specifications, and (2) a code synthesis approach that expandseach function recursively from the generated specifications. We conductextensive evaluation on various robot task benchmarks, including a novel gamebenchmark Robotouille, designed to simulate diverse cooking tasks in a kitchenenvironment. The project's website is available athttps://portal-cornell.github.io/demo2code/

Demo2Code: From Summarizing Demonstrations to Synthesizing Code via Extended Chain-of-Thought