Embeddings & Similarity
LLMs turn text into embeddings β vectors of numbers. Similar meanings have vectors pointing the same way, measured by cosine similarity:
cosine(a, b) = dot(a, b) / (|a| Γ |b|)
Result ranges from 1 (identical direction) to -1 (opposite direction), with
0 meaning unrelated. In practice, text embedding vectors are non-negative, so
similarity scores land in [0, 1]. This is the math behind semantic search and RAG.
Your task
Write cosine(a, b) for two equal-length vectors, rounded to 2 decimals.
β‘ 45 XP
solution.jsRuns in your browser
Loading editorβ¦
Tests3 to pass
identical direction β 1
orthogonal β 0
[3,4] vs [4,3] β 0.96
Run your code to check these.