3 ms·
Started hearing about it in ~2022 when some ML researchers accidentally left a model training on over a weekend. For a while the model wasn’t doing much (so the
by janalsncm 1y ago
Started hearing about it in ~2022 when some ML researchers accidentally left a model training on over a weekend. For a while the model wasn’t doing much (so they were going to turn it off) and then over the weekend it got surprisingly good.
https://en.m.wikipedia.org/wiki/Grokking_(machine_learning) https://en.m.wikipedia.org/wiki/Grokking_(machine_learning)