Back to Learn
Intermediate

Sub-Query Generation

Sub-query generation is a query transformation strategy where an LLM decomposes a complex user prompt into multiple, simpler atomic questions to enable multi-hop reasoning and parallelized retrieval. While it significantly improves recall for multi-faceted questions, it introduces higher latency and increased token costs due to multiple LLM calls.

Definition

Sub-query generation is a query transformation strategy where an LLM decomposes a complex user prompt into multiple, simpler atomic questions to enable multi-hop reasoning and parallelized retrieval. While it significantly improves recall for multi-faceted questions, it introduces higher latency and increased token costs due to multiple LLM calls.

Disambiguation

Distinguish from SQL subqueries; this is natural language decomposition for semantic search, not relational database nesting.

Visual Metaphor

"A project manager breaking a massive 'Launch Product' task into specific sub-tickets for different departments to handle simultaneously."

Conceptual Overview

Sub-query generation is a query transformation strategy where an LLM decomposes a complex user prompt into multiple, simpler atomic questions to enable multi-hop reasoning and parallelized retrieval. While it significantly improves recall for multi-faceted questions, it introduces higher latency and increased token costs due to multiple LLM calls.

Disambiguation

Distinguish from SQL subqueries; this is natural language decomposition for semantic search, not relational database nesting.

Visual Analog

A project manager breaking a massive 'Launch Product' task into specific sub-tickets for different departments to handle simultaneously.

Related Articles