Efficient Code Embeddings 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 recipe and demonstrate state-of-the-art performance despite the relatively small size of the models, validating this approach to code embedding model construction.

9 pages. Accepted at the NeurIPS 2025 Workshop on Deep Learning for Code (DL4CODE)

Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)

arXiv:2508.21290 [cs.CL]

(or arXiv:2508.21290v2 [cs.CL] for this version)

https://doi.org/10.48550/arXiv.2508.21290 Focus to learn more arXiv-issued DOI via DataCite

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.