LlamaIndex Quiz

Answers and explanations appear as you take the quiz.

1. What is LlamaIndex primarily designed to do?
   • Connect LLMs to your data for retrieval-augmented generation
   • Train new foundation models
   • Generate images and video
   • Manage GPU clusters in the cloud
2. What was LlamaIndex originally called when it launched?
   • DataChain
   • GPT Index
   • VectorBase
   • LlamaHub
3. Why is retrieval needed instead of just pasting everything into the prompt?
   • LLMs cannot read text files
   • Embeddings are illegal to store
   • Models have limited context windows and no knowledge of private data
   • Retrieval makes models train faster
4. In a LlamaIndex pipeline, what is created from each chunk of a document to enable semantic search?
   • A compressed ZIP file
   • A SQL table row
   • A screenshot
   • A vector embedding
5. What is LlamaParse built to handle?
   • Parsing complex PDFs, tables, and documents for ingestion
   • Parsing source code into syntax trees
   • Parsing audio into transcripts
   • Parsing network packets
6. Where do LlamaIndex node postprocessors run in the documented query pipeline?
   • Before documents are parsed or indexed
   • After retrieval and before response synthesis
   • Only after a final answer is returned
   • Between model pretraining and tokenizer training
7. What does LlamaIndex’s SubQuestionQueryEngine do with a complex multi-source question?
   • Converts every document into a single unsearchable string
   • Runs only a fixed keyword lookup with no query planning
   • Creates subquestions for relevant data sources and synthesizes the intermediate responses
   • Fine-tunes the language model on the documents
8. In a LlamaIndex query pipeline, what is the response synthesizer responsible for?
   • Choosing the original PDF files before ingestion
   • Training the embedding model from scratch
   • Replacing the retriever’s index with a file system
   • Combining the query and retrieved Nodes to produce a response