Efficient Code Embeddings from Code Generation Models

Daria Kryvosheieva, Saba Sturua, Michael Günther, Scott Martens, Han Xiao

Keywords: code embeddings; retrieval; embedding models

TL;DR: Two new code embeddings with SOTA performance, derived from code generation models

Abstract: jina-code-embeddings is a novel code embedding model suite designed to retrieve code from natural language queries, perform technical question-answering, and identify semantically similar code snippets across programming languages. It makes innovative use of an autoregressive backbone pre-trained on both text and code, generating embeddings via last-token pooling. We outline the training methodology and demonstrate state-of-the-art performance despite the relatively small size of the models, validating this approach to code embedding model construction.

Submission Number: 52