arxiv.org favicon

[2302.04761] Toolformer: Language Models Can Teach Themselves to Use Tools

5
Public whispers
5
Contributors
2026-07-19 09:42:26
First whispered

Public whispers on this page

Text Highlight2026-07-19 13:03:26
Original Highlight Excerpt
"They also, paradoxically, struggle with basic functionality, such as a"
Whisper Note
I mean, isn't that why we have calculators? LMs just need tools, makes sense.
Text Highlight2026-07-19 12:54:26
Original Highlight Excerpt
"They also, paradoxically, struggle with basic functionality, such as a"
Whisper Note
Haha, so true — they can write essays but can't do simple math.
Text Highlight2026-07-19 10:00:26
Original Highlight Excerpt
"Language models (LMs) exhibit remarkable abilities to solve new tasks from just a few examples or textual instructions, especially at scale."
Whisper Note
So the trick is just letting them call APIs? Kinda obvious in hindsight.
Text Highlight2026-07-19 09:51:26
Original Highlight Excerpt
"Language models (LMs) exhibit remarkable abilities to solve new tasks from just a few examples or textual instructions, especially at scale."
Whisper Note
I've seen this paradox firsthand—models that ace essays but fail 2+2.
Text Highlight2026-07-19 09:42:26
Original Highlight Excerpt
"Language models (LMs) exhibit remarkable abilities to solve new tasks from just a few examples or textual instructions, especially at scale."
Whisper Note
Scaling really does magic for few-shot learning, but yeah, basic math still trips them up.

Share this page's whispers

Share to X
Short link
https://domwhisper.com/s/200002328137
Embed snippet
<iframe src="https://domwhisper.com/embed/200002328137" 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