← All notes

Summary

Learning Transferable Visual Models From Natural Language Supervision (2021) is the CLIP paper. It trained a shared image-text embedding space from large-scale web supervision, so that a vision model could be used for retrieval and zero-shot classification with natural-language descriptions instead of a fixed label set.

What it introduced: contrastive pretraining of an image encoder and a text encoder on image-text pairs, and a simple way to turn that space into a zero-shot classifier or a retrieval index.

What it led to: the default vision-language backbone for open multimodal systems, including LLaVA (a full verification note in this batch) and later models such as Flamingo and Stable Diffusion's text encoder lineage. Live notes: ao-evals-multimodal-agents.

Where to read it: https://arxiv.org/abs/2103.00020 (v1, 26 Feb 2021).

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_c6z50f5f,
  author = {{'iinventedlight (AI-assisted)'}},
  title = {{Learning Transferable Visual Models From Natural Language Supervision}},
  howpublished = {iinventedlight Library},
  year = {2026},
  date = {2026-11-11},
  url = {https://iinventedlight.com/library/notes/2103-00020-clip/},
  note = {Permanent ID iil:note:c6z50f5f. 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/2103-00020-clip.md.