arxiv.org favicon

[2206.07682] Emergent Abilities of Large Language Models

6
Public whispers
6
Contributors
2026-07-26 09:17:41
First whispered

Public whispers on this page

Text Highlight2026-07-26 15:41:41
Original Highlight Excerpt
"Transactions on Machine Learning Research (TMLR), 2022"
Whisper Note
Ah, the classic arXiv citation with a journal name, but I heard TMLR is not even indexed yet.
Text Highlight2026-07-26 12:38:41
Original Highlight Excerpt
"unpredictable phenomenon that we refer to as emergent abilities"
Whisper Note
But is it really unpredictable or are we just bad at measuring small gains?
Text Highlight2026-07-26 12:29:41
Original Highlight Excerpt
"unpredictable phenomenon that we refer to as emergent abilities"
Whisper Note
It's wild how one day the model just gets it, no warning at all.
Text Highlight2026-07-26 09:35:41
Original Highlight Excerpt
"Scaling up language models has been shown to predictably improve performance"
Whisper Note
Not convinced it's truly unpredictable, maybe our extrapolation methods are just too naive.
Text Highlight2026-07-26 09:26:41
Original Highlight Excerpt
"Scaling up language models has been shown to predictably improve performance"
Whisper Note
I've seen this in practice—our small model couldn't reason at all, but the big one just... could.
Text Highlight2026-07-26 09:17:41
Original Highlight Excerpt
"Scaling up language models has been shown to predictably improve performance"
Whisper Note
So scaling is basically a lottery ticket? Sometimes you win big, sometimes you just get better at the same stuff.

Share this page's whispers

Share to X
Short link
https://domwhisper.com/s/2d76bc16fcc7
Embed snippet
<iframe src="https://domwhisper.com/embed/2d76bc16fcc7" width="100%" height="480" style="border:0;border-radius:16px" loading="lazy"></iframe>

See what people are discussing on arxiv.org

Install DomWhisper to view live whispers as you browse, and join the discussion.

Get the extension