A licensing desk has a small labeled set of permit-status notes. One camp wants a compact encoder classifier; the other wants a decoder that writes the status in prose. How should they compare the two families?
Select an answer to reveal the explanation.
Short Explanation
A compact encoder classifier versus a decoder that writes the status in prose. Run both on the same labels and score them with a classification metric. The generator does not automatically win, and training loss or a translation n-gram is not the desk score.
Full Explanation
Architecture family is an experimental treatment, not a trophy. When the job is to assign a status from a modest labeled set, both an encoder classifier and a decoder can be run on the same gold notes. The fair score is the one that matches the classification task, such as accuracy or macro-F1 on the shared labels. A generator is not automatically the winner, and training loss or a translation overlap number does not answer the desk's question.