Researchers at the University of Southern California have developed ProgPrompt, where the LLM is given prompts in Python that include a sample question and solution to constrain its responses to the tasks the robot can carry out.
Google researchers have come up with a tactic involving the use of a list of executable behaviors for robots with the PaLM LLM. This model responds to human inquiries with conversational language and selects a behavior from the predetermined list.
Collection
[
|
...
]