Google Search

• LLM in the loop with preference optimization


• synthetic data generation


• cross modality "distillation" / dictionary remapping


• constrained decoding

Blended search is already with us, and will be refined and developed over the coming years; search engines will continue to hone their ability to understand context in search queries – to divine exactly what users are looking for and to deliver ever more relevant results.

By grounding LLMs with use-case-specific information through RAG, the quality and accuracy of responses are improved.