RAG(Retrieval-Augmented Generation) — Learn free

Self-check

Questions 1–10 of 10

1. What problem with large language models (LLMs) does Retrieval-Augmented Generation (RAG) primarily address?
Question 1A. LLMs often generate incorrect answers with confidence, especially about recent events or niche topicsB. LLMs are too slow to type responsesC. LLMs cannot process text longer than one sentenceD. LLMs refuse to answer any questionE. LLMs can only work with numeric data
2. What are the two fundamental components of a RAG system?
Question 2A. A tokeniser and a compilerB. A retriever and a generatorC. A database and a spreadsheetD. An encoder and a firewallE. A cache and a scheduler
3. Which paper first introduced the retrieve-then-generate pattern that RAG follows?
Question 3A. Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (Lewis et al., 2020)B. Learning Transferable Visual Models (Radford et al., 2021)C. Reading Wikipedia to Answer Open-Domain Questions (Chen et al., 2017)D. Attention Is All You NeedE. The DeepSeek-V3 technical report
4. In the four phases of how RAG works, what happens during context construction?
Question 4A. The user's query is deletedB. The model is retrained from scratchC. Citations are validated after the answer is producedD. The retrieved documents or passages are formatted and added to the model's promptE. The vector database is rebuilt
5. Which chunking strategy first splits documents into sections, then paragraphs, and finally into sentences if necessary?
Question 5A. Fixed-length chunkingB. Token-based chunkingC. Domain-specific chunkingD. Contextual chunkingE. Recursive chunking
6. In the query rewriting example, the user input 'How about Mustafa?' is rewritten to 'When was the last time Mustafa made a purchase from us?'. What does this illustrate?
Question 6A. Clarifying an ambiguous query so that retrieval matches documents more accuratelyB. Compressing a long query into a short oneC. Translating a query into another languageD. Removing personal data from a queryE. Caching a frequently asked question
7. Which advanced RAG technique converts both queries and documents into vector representations so that semantically relevant documents are found even without exact keyword matches?
Question 7A. Query expansionB. Dense retrievalC. Modular RAGD. CachingE. Time-based reranking
8. How does Agentic RAG differ from standard RAG?
Question 8A. It only works with imagesB. It removes the generator componentC. It actively determines how and where to search, dynamically adjusting its strategy, instead of passively following predefined retrieval stepsD. It requires a human to approve every retrievalE. It never uses external data
9. Which of the following is NOT one of the four critical security considerations for RAG systems?
Question 9A. Prevent embedding of PII or sensitive data in vector databasesB. Implement strict access controls for vector databasesC. Secure access to LLM APIsD. Disable all logging of API interactionsE. Validate generated data before delivering responses to clients
10. In evaluating RAG systems, which metric examines how closely the generated response adheres to the retrieved information without introducing hallucinations?
Question 10A. FluencyB. RecallC. PrecisionD. LatencyE. Faithfulness