Definition
The iterative process where an AI agent or RAG pipeline resolves complex queries by sequentially retrieving and synthesizing information across multiple disparate data sources. It involves using the context gathered from an initial retrieval step to formulate subsequent queries, effectively bridging information gaps that a single-pass retrieval cannot address.
Unlike parallel retrieval which fetches multiple chunks at once, multi-hop reasoning is sequential and dependent on previous outputs.
"A scavenger hunt where each found clue provides the specific directions needed to find the next location."
Conceptual Overview
The iterative process where an AI agent or RAG pipeline resolves complex queries by sequentially retrieving and synthesizing information across multiple disparate data sources. It involves using the context gathered from an initial retrieval step to formulate subsequent queries, effectively bridging information gaps that a single-pass retrieval cannot address.
Disambiguation
Unlike parallel retrieval which fetches multiple chunks at once, multi-hop reasoning is sequential and dependent on previous outputs.
Visual Analog
A scavenger hunt where each found clue provides the specific directions needed to find the next location.