A research team develops a method to summarize lengthy documents by creating a compact, continuous representation that allows a language model to answer questions about the original document without re-reading it. Another team uses a very similar method to create a compact, continuous representation that guides a model to adopt a specific persona (e.g., a pirate) based on a few examples of pirate speech. Analyze the underlying principle that connects these two seemingly different applications. In your analysis, explain how one of these concepts can be viewed as a specific instance of the other.