Text and code embeddings by contrastive pre-training

In three linesOpenAI releases a contrastive pre-training method for generating text and code embeddings. The technique improves semantic representation of data for search and similarity tasks.

Summary generated by Claude — human-verified