BtwIUseSystemd's profile picture.

oooo

@BtwIUseSystemd

oooo reposted

LLMs can win Maths Olympiad, yet still fail to answer dumb questions like "which number is bigger, 9.11 or 9.9?" @karpathy called it Jagged Intelligence. Research by MIT & Berkeley found that zeroing out the "Bible verse neurons" improved 9.8 vs 9.11 accuracy by 21% in llama-3.1…

Yuchenj_UW's tweet image. LLMs can win Maths Olympiad, yet still fail to answer dumb questions like "which number is bigger, 9.11 or 9.9?" @karpathy called it Jagged Intelligence.

Research by MIT & Berkeley found that zeroing out the "Bible verse neurons" improved 9.8 vs 9.11 accuracy by 21% in llama-3.1…

Jagged Intelligence The word I came up with to describe the (strange, unintuitive) fact that state of the art LLMs can both perform extremely impressive tasks (e.g. solve complex math problems) while simultaneously struggle with some very dumb problems. E.g. example from two…

karpathy's tweet image. Jagged Intelligence

The word I came up with to describe the (strange, unintuitive) fact that state of the art LLMs can both perform extremely impressive tasks (e.g. solve complex math problems) while simultaneously struggle with some very dumb problems.

E.g. example from two…
karpathy's tweet image. Jagged Intelligence

The word I came up with to describe the (strange, unintuitive) fact that state of the art LLMs can both perform extremely impressive tasks (e.g. solve complex math problems) while simultaneously struggle with some very dumb problems.

E.g. example from two…
karpathy's tweet image. Jagged Intelligence

The word I came up with to describe the (strange, unintuitive) fact that state of the art LLMs can both perform extremely impressive tasks (e.g. solve complex math problems) while simultaneously struggle with some very dumb problems.

E.g. example from two…
karpathy's tweet image. Jagged Intelligence

The word I came up with to describe the (strange, unintuitive) fact that state of the art LLMs can both perform extremely impressive tasks (e.g. solve complex math problems) while simultaneously struggle with some very dumb problems.

E.g. example from two…


United States Trends

Loading...

Something went wrong.


Something went wrong.