A public-pool desk has a dozen adaptation attempts sitting in chat screenshots and cannot match which learning rate produced which score. What should they keep instead?
Select an answer to reveal the explanation.
Short Explanation
Chat screenshots split knobs from scores. Keep one experiment log that stores hyperparameters, data identity, and metrics together for each run. A named vendor tracker is not required. Kernels, empty ops tickets, and the winning score alone are not a notebook.
Full Explanation
A run log is the experiment’s lab notebook: knobs, data identity, and metrics for each attempt live in one place. Chat screenshots split those fields apart, so a learning rate cannot be matched to a score. The official exam does not name a vendor tracker; a generic log is enough. Kernels, empty ops tickets, and score-only scraps are not that notebook.