A city archives desk wants a language model to draft finding-aid blurbs for newly accessioned boxes. A volunteer is already launching overnight adaptation jobs just to see what happens. What must be written down before the first run?
Select an answer to reveal the explanation.
Short Explanation
Before the first overnight job, write the claim and the win bar. Adaptation is an experiment, not an open-ended training loop. A CUDA topology, a leaked test peek, and a datacenter ticket are not that design step.
Full Explanation
An LLM adaptation is an experiment, so the claim under test and the success criterion belong on paper before any job is launched. The hypothesis states what change is expected to improve the blurbs; the success bar is the number or rubric that would count as a win. Overnight jobs without those two items cannot tell a finished run from an open-ended training loop. Hardware topology, leaked test scores, and datacenter queues are not substitutes for that design step.