Vector Embeddings
HoldTechniques
Numerical representations that capture semantic similarity between texts or labels.
Why it's here
Placed in Hold: 1 article(s) of evidence from 1 source(s), led by research-stage coverage, with 0 in the last 30 days. Confidence 24%. Low accumulated evidence, so it defaults conservatively pending more signal.
Evidence (1)
- 4Simon Willison·8/14/2026researchGenerating Tags with Hallucinated Labels
Simon Willison describes a tagging workflow for older blog content where an LLM generates likely classifications without being given the full existing tag set. The imagined tags are then matched against the current vocabulary using vector embeddings, avoiding the need to load thousands of tags into the prompt at once. The post cites Doug Turnbull’s approach as a practical way to improve tag suggestions for large taxonomies.