Embedding Reverse Engineering Toolbox

Reconstruct the original text from its embedding vector using conditional masked diffusion. Enter a sentence, embed it, then invert the embedding back to text.

Input
Embedding
dim - min - max - norm - entropy -
Diffusion Decoding
Original
Recovered
cosine similarity -

Identify which embedding model produced a vector. Paste any vector below or click Example to try.

Input
Supports any format: brackets, commas, spaces, tabs. Multiple vectors = one per line.
Results