← All notes

Summary

Toolformer: Language Models Can Teach Themselves to Use Tools (2023) showed that a language model can learn when to call external tools — such as a calculator, a search API, or a calendar — by self-supervised training, rather than only by hand-written tool prompts.

What it introduced: a method for inserting tool calls into a model's own text, keeping the calls that actually help next-token prediction, and then fine-tuning on that self-labeled data so the model decides when to use a tool.

What it led to: a large part of the later tool-use and function-calling literature, including agents that learn or make their own tools. Live notes: ao-tools-llms-as-tool-makers. For the security side of tool-using systems, see the indirect prompt-injection pointer in this batch and ao-security-prompt-injection-design.

Where to read it: https://arxiv.org/abs/2302.04761 (v1, 9 Feb 2023).

The paper is posted on arXiv under the arXiv non-exclusive distribution license. This note links to it rather than reproducing it. The note itself is CC BY 4.0.

License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Pointer by iinventedlight (AI-assisted). Paper license: arXiv non-exclusive distribution; this note links to it and does not reproduce it.

@misc{iil_note_nej2ska0,
  author = {{'iinventedlight (AI-assisted)'}},
  title = {{Toolformer: Language Models Can Teach Themselves to Use Tools}},
  howpublished = {iinventedlight Library},
  year = {2026},
  date = {2026-11-11},
  url = {https://iinventedlight.com/library/notes/2302-04761-toolformer/},
  note = {Permanent ID iil:note:nej2ska0. License: CC-BY-4.0.},
}

Markdown

Spotted a mistake? Use the MCP feedback tool or email hello@iinventedlight.com.

For language models and agents: start at /llms.txt | /agents.md | /.well-known/agent-library.json. This page in Markdown: /library/notes/2302-04761-toolformer.md.