1Z0-184-25 Question 60
Single answerYou have configured an IVF Vector Index to speed up semantic search within an Oracle Cloud Infrastructure 2025 application. Your users report that while query performance is good, many relevant documents do not appear in the results. Which action would most likely improve the recall of your search without unraveling the performance benefits of the IVF Vector Index?
- A
Add more centroid clusters to the IVF Vector Index definition.
- B
Reduce the dimensionality of your vector embeddings by discarding less significant components.
- C
Increase the search parameter so that more candidate clusters are scanned at query time.
- D
Store the IVF Vector Index on higher-performance block volumes.
Show answer and explanation
Correct answer: C
Explanation
In an IVF Vector Index, query-time parameters�often controlling how many clusters or cells are probed�significantly affect recall. By scanning more clusters, you reduce the chance of missing relevant embeddings. For additional guidance, refer to official Oracle Cloud Infrastructure documentation on Vector Search, which explains how adjusting the search parameter offers a balanced approach between recall and efficiency.
- A. Incorrect.
Option 1: Although adding more centroid clusters can increase the granularity of the index, it often makes index construction more expensive and complex. This change alone may not necessarily improve recall unless you also modify the query-time parameters to search a greater number of clusters.
- B. Incorrect.
Option 2: Reducing dimensionality can lessen storage and computational overhead, but it may also decrease accuracy and recall because vital information in the vector embeddings might be lost in the process.
- C. Correct.
Option 3: Increasing the search parameter ensures more centroid clusters are scanned at query time, which generally leads to higher recall. This parameter controls how thoroughly the IVF index searches within the nearest clusters before returning results.
- D. Incorrect.
Option 4: Using higher-performance storage can speed up I/O but does not address the core issue of insufficient coverage of relevant documents during the search process.