Precise Zero-Shot Dense Retrieval without Relevance Labels (HyDE)

Instructed an LLM to generate a hypothetical answer document to an input query, then embedded the hypothetical document to search the corpus in document-document embedding space.

Directly guides architectural decisions, alignment strategy, and serving infrastructure for task-question-answering, task-feature-extraction.

Assumptions

• Empirical distribution regularity holds and target domain adheres to pretraining linguistic/visual support

Limitations

• Resource scaling, inference memory requirements, and alignment robustness vary with model size and hardware topology