What’s on the Embeddings poster
A definition and 6 facts worth remembering.
A model that maps a piece of text to a list of numbers — a vector — positioned so that texts with similar meanings sit close together in that space.
- Typical sizes run from 384 to 3,072 dimensions.
- Similarity is usually cosine distance between vectors.
- Two paraphrases land close together even with no shared words.
- Used for search, clustering, deduplication and recommendation.
- The same model must embed both query and documents.
- Changing embedding model means re-embedding everything.
- Common size — 768–1536
- Metric — cosine
- Range — −1 to 1
- 'How do I reset my password' matches a document titled 'Account recovery' with no wor…
Embeddings capture similarity, not truth. Two confidently opposite statements can sit very close together.
Questions about the Embeddings poster
- What’s on the Embeddings poster?
- A definition and 6 facts worth remembering. A model that maps a piece of text to a list of numbers — a vector — positioned so that texts with similar meanings sit close together in that space. Typical sizes run from 384 to 3,072 dimensions.; Similarity is usually cosine distance between vectors.; Two paraphrases land close together even with no shared words.; Used for search, clustering, deduplication and recommendation.; The same model must embed both query and documents.; Changing embedding model means re-embedding everything.; Common size — 768–1536; Metric — cosine; Range — −1 to 1; 'How do I reset my password' matches a document titled 'Account recovery' with no wor…. Embeddings capture similarity, not truth. Two confidently opposite statements can sit very close together.
- Who is the Embeddings poster for?
- Embeddings belongs to the Computing section rather than to a school year, because computing is not something one grade owns. Anyone learning ai & ml can pin it up — a beginner, a student mid-course, or someone revising years later.
- When should you use the Embeddings poster?
- Similar text lands near similar text. Everything embeddings are used for follows from that. A wall chart earns its place by being glanceable from where the work is happening, so Embeddings belongs on the wall where that computing work actually happens, within glancing distance, rather than filed away.
- What other posters go with Embeddings?
- Evaluating AI Output, LLM Vocabulary and Kinds of Model sit alongside Embeddings in the Computing section. Printed together they make a wall rather than a single sheet, which is how a reference set actually gets used.Evaluating AI OutputLLM VocabularyKinds of Model
- Is the Embeddings poster free to download and print?
- Yes. Embeddings downloads as a free PDF with no account, no email and no watermark, like everything else in the Computing section. Print as many copies as you like for a home, a classroom, a library or a tutoring group; reselling the file is the only thing the licence rules out.Read the licence
- What size does the Embeddings poster print at?
- Embeddings is a vector PDF laid out for US Letter, and prints on A4 with Fit to page — the same file, no separate download. Because every mark on it is drawn rather than photographed, it stays sharp enlarged to A3, A2 or A1 at a copy shop. Colour carries emphasis only, so a greyscale print of Embeddings loses nothing.Printing guide
Related posters
Charts that sit alongside Embeddings on the same wall.
Browse every Computing poster, or start from the full catalogue.