“And yeah, so this is a downstream evaluation. This is the…” — Latent Space: The AI Engineer Podcast at 17:12
“And yeah, so this is a downstream evaluation. This is the human evaluation, the human eval, negative likelihood, and basically corresponds to the performance. So we're basically perfectly able to match after just two hours of training, the downstream performance on these tasks, and then continued training allows it…”
0:00
32:59