🪆 Matryoshka Embedding Explorer

Built with anycoder

1. Input Text

Initializing model...

3. Similarity Retention Analysis

This chart shows the cosine similarity between the full 1024D embedding and its truncated (zero-padded) representation at various dimensions. Higher similarity indicates better information retention.