What is an embedding vector?

An embedding vector is a fixed-length list of floating-point numbers that represents a piece of data — a word, sentence, image, or any other input — in a high-dimensional mathematical space. Machine learning models such as BERT, OpenAI's text-embedding models, and image encoders produce these vectors so that semantically similar items end up geometrically close together. A single embedding can have anywhere from a few dozen dimensions to several thousand.

Understanding the numerical properties of an embedding is useful when debugging models, comparing vector representations, detecting anomalies, or optimizing storage and retrieval in a vector database.

Tool description

This tool accepts a JSON array of numbers representing an embedding vector and instantly computes a set of descriptive statistics: the number of dimensions, minimum and maximum values, arithmetic mean, standard deviation, L2 norm (magnitude), and sparsity. It also surfaces the top 10 most influential dimensions ranked by absolute activation value.

Features

  • Instant statistics — dimensions, min, max, mean, standard deviation, L2 norm, and sparsity computed in the browser without sending data to a server.
  • Top 10 activations table — lists the ten dimensions with the largest absolute values, sorted by impact, with their index, raw value, and absolute value.
  • JSON editor with syntax highlighting — paste or type any valid JSON array directly into the editor and get results in real time.