nostr:npub18k4cppxhna6hq8p7hq26k6yfj00wt8040f8d47k7qutc4zvvtrps6apvgt 
To be clear, greedy sampling is according to presenters at the conference I'm attending is used a lot in business applications. Profitable versus probable seems like semantics. Are you perhaps describing the role of prompt tuning to provide better answers?  My little definition was intended as the simplest definition that describes the process for writers, and how it was described to me by IBMers building the company's AI tech, but keep in mind it was introductory.