RAG & Vector Databases test (Advanced)

Sample questions

5 sample questions from the RAG & Vector Databases test, with the answers. Samples are taken out of the live test, so you will not meet these when you sit it.

1. Which of the following statements is true regarding the impact of chunking strategies on RAG systems?

   1. Chunking strategies have no impact on retrieval performance since vector similarity metrics are entirely independent of how text is divided into chunks.
   2. Chunking strategies only influence the computational cost of embedding generation, not retrieval performance.
   3. Chunking strategies significantly affect retrieval performance by influencing how well the vector database can match queries to relevant document sections.
   4. Chunking strategies primarily affect the size of the vector database but not the quality of retrieval results.

   Answer: C. Chunking strategies significantly affect retrieval performance by influencing how well the vector database can match queries to relevant document sections.

   Chunking strategies define how text is divided into smaller segments before being embedded and stored in a vector database. The choice of chunk size and overlap can impact retrieval performance, as larger chunks may capture more context but may also dilute the relevance of specific information, while smaller chunks may improve precision but could miss broader context. Overlapping chunks can help maintain continuity and reduce information loss.

2. Which of the following is a key aspect of chunking strategies in RAG systems that ensures better context capture between adjacent text segments?

   1. Embedding model alignment
   2. Chunk overlap strategies
   3. Vector similarity metrics
   4. Query expansion techniques

   Answer: B. Chunk overlap strategies

   Chunk overlap strategies are a critical component of chunking strategies in RAG systems. They ensure that overlapping chunks capture context better, reducing the risk of missing important information that lies at the boundaries of chunks. This is especially relevant for tasks like question answering, where context is vital. Other options may be related but do not directly align with the concept being tested here.

3. You are building a RAG system using LlamaIndex and need to extract metadata like document source, section, and date for filtering queries. How can you accomplish this in LlamaIndex?

   1. Use `MetadataFilter` as a separate component to extract metadata and pass it to the `VectorStoreIndex`
   2. Attach metadata to the `Document` objects and use `text_splitter` to extract it during indexing
   3. Use the `metadata` parameter when creating the `VectorStoreIndex` and query with `filter` in `as_query_engine()`
   4. Use `NodeParser` to extract metadata from the document's text and store it in the `VectorStoreIndex`

   Answer: C. Use the `metadata` parameter when creating the `VectorStoreIndex` and query with `filter` in `as_query_engine()`

   LlamaIndex allows metadata to be stored with each document and indexed in the `VectorStoreIndex`. To use metadata during querying, you can specify filters using the `filter` parameter in `as_query_engine()`. This enables efficient metadata-based retrieval.

4. Which of the following statements correctly describes a common consideration when implementing chunk overlap strategies in RAG systems?

   1. Overlapping chunks can help retain context but may lead to redundant information if not managed properly.
   2. Chunk overlap is unnecessary in RAG systems as it increases computational complexity without any benefits.
   3. Chunk overlap is only relevant for text documents, not for image or audio data.
   4. Chunk overlap should always be set to 50% of the chunk size for optimal results.

   Answer: A. Overlapping chunks can help retain context but may lead to redundant information if not managed properly.

   Chunk overlap strategies are important in RAG systems to maintain context between adjacent chunks. However, if the overlap is too large, it can result in redundant information being stored and queried, which may impact the efficiency and performance of the system. The other options either describe unrelated concepts or are incorrect statements about chunking.

5. What is the primary purpose of chunk overlap strategies in RAG systems?

   1. Chunk overlap strategies are primarily used to reduce the computational cost of generating embeddings by reusing previous chunks.
   2. Chunk overlap strategies are used to increase the size of the vector database by creating more unique embeddings for each chunk.
   3. Chunk overlap strategies help reduce information loss by allowing overlapping segments of text to be indexed, ensuring that context around boundaries is preserved during retrieval.
   4. Chunk overlap strategies are designed to ensure that all chunks are of uniform length, which simplifies the retrieval process.

   Answer: C. Chunk overlap strategies help reduce information loss by allowing overlapping segments of text to be indexed, ensuring that context around boundaries is preserved during retrieval.

   Chunk overlap strategies are used in RAG systems to ensure that the context around the boundaries of text chunks is preserved, which helps in reducing information loss and improving retrieval accuracy. This is achieved by having overlapping segments of text, which allows the model to better understand the context of the query. The other options may describe benefits of chunking or other strategies, but they do not directly address the purpose of chunk overlap.