https://vubon.dev/tags/algorithms/Recent content in Algorithms on Vubon NotesHugoen-usSun, 04 Oct 2026 12:00:00 +0700https://vubon.dev/posts/understanding-cosine-similarity-in-vector-search/Sun, 04 Oct 2026 12:00:00 +0700https://vubon.dev/posts/understanding-cosine-similarity-in-vector-search/<p>If you have spent any time working with Large Language Models, Retrieval-Augmented Generation (RAG), or vector databases, you have seen this term everywhere:</p> <p><strong>Cosine Similarity.</strong></p> <p>When you ask an AI assistant a question, the system converts your text into a list of numbers (an embedding), compares it against millions of stored vectors, and returns the most relevant chunks.</p> <p>Almost every vector database tells you: <em>&ldquo;We use cosine similarity to rank semantic relevance.&rdquo;</em></p>