Hypothetical Document Embeddings (HyDE)

2 phases system:

1. Generate a hypothetical answer (from user query):
   • Something like write a document to answer {user_query}, utilizing a LLM.
   • The answer may not be factually accurate, but it serves as an example of what a relevant document could look like.
2. Encoding and Retrieval:
   • The hypothetical answer is then processed by an unsupervised contrastive encoder
   • Retrieve most similar document to the hypothetical answer’s embedding.