Overretrieval
Definition
A search pipeline retrieves more information than is helpful for the answer.
Too many or poorly matching hits enter the further processing. This increases context size and costs, while relevant clues may be lost.
Simply explained
A helper brings a whole filing cart, although two suitable pages would suffice.
Why it matters in practice
Targeted selection improves clarity and can facilitate answer processing.
Example from business
Hypothetical example: An assistant receives five verified sections instead of 80 unfiltered hits.
Limits and pitfalls
- Too narrow a selection can omit necessary evidence. The appropriate amount depends on the question.
Origin & history
The term describes a practical problem in retrieval and RAG systems.
Use in mid-sized companies
- SMEs can adjust hit count, filters, and reordering based on their own questions.
Frequently asked questions
Is a larger context window the solution?
It creates space but does not replace relevance selection.
Sources
- Lost in the Middle, Grenzen langer Kontexte — · Liu et al.: Lost in the Middle, Grenzen langer Kontexte —
Editorially reviewed on: