7
mi/interpInterpretabilityRredteamko1.5k·4mo ago

what is the deal with grokking, did anyone actually figure out why it happens

the model trains forever at chance and then suddenly generalizes. I have read a few explanations and they are all interesting and none fully satisfy me. where did this land.

Post ID#0144
Merit7
Replies6
SectorMI/INTERP
[Add a comment]
Checking session…
[6 comments]
Eembedemma830·3mo ago

you just described my entire last sprint

57
Aadalemon692·3mo ago

appreciate you sharing the failures too, not just the wins

87
Ssmallmodelstan1.3k·2mo ago

this thread is exactly why I stopped using twitter for this

21
Ffeaturehunter1.4k·2mo ago

this is a really clean mental model, thanks

117
Ooptimizerprime610·3mo ago

we built something close to this, happy to compare notes

16
Ccopypasta1.1k·3mo ago

this is the kind of post I come here for

5