← Back to Quiz Home
This quiz covers the mechanics of RAG (Retrieval-Augmented Generation), including embeddings, chunking, and agents.
What is "Chunking" in the context of Knowledge Bases?
Chunking is critical because LLMs have a fixed context window. Sending a whole book is expensive; creating chunks retrieval more precise.
Which chunking strategy splits text where the meaning changes (e.g., between distinct topics) rather than just by token count?
Semantic chunking uses an embedding model to determine breakpoints based on topic shifts.
What are Agents for Amazon Bedrock?
Agents can "Reason" -> "Act" -> "Observe" to solve problems like "Book a flight and email me the receipt".
How does an Agent know which external API to call?
The OpenAPI schema (Swagger) serves as the "instruction manual" for the LLM to use your tools.
What is "Chain-of-Thought" (CoT) prompting?
CoT significantly improves performance on complex math or logic problems.
Which metric evaluates whether the RAG answer is derived only from the retrieved context (preventing hallucinations)?
Faithfulness measures if the claims in the answer can be inferred from the context provided.
Hybrid search leverages the best of both worlds: exact matching for unique IDs and semantic matching for concepts.
What is "Hierarchical Chunking"?
This strategy helps maintain the broader context (Parent) while allowing the search to pinpoint specific details (Child).
Which AWS service provides the "Thought Trace" (CoT) logs for Bedrock Agents?
You can view the agent's "Pre-computation", "Invocation", and "Post-computation" steps to debug its reasoning.
When would you use Provisioned Throughput in Bedrock?
"On-demand" has shared limits; Provisioned Throughput reserves dedicated compute for your model.
What is the role of an "Action Group" in Bedrock Agents?
Action Groups bridge the gap between the LLM's text output and actual code execution (Lambda).
What is "Embeddings" in GenAI?
"King" - "Man" + "Woman" ≈ "Queen" is the classic example of vector math on embeddings.
Which component is responsible for retrieving relevant documents in a RAG system?
The Retriever scans the vector index to find chunks most similar to the user's query.
How do you handle a user request that requires data from a private SQL database using Bedrock?
Agents allow you to write a Lambda function that executes the SQL query securely and returns the result to the LLM.
What is "Context Precision" in RAG evaluation?
High precision means the retriever is finding the right documents, not just random ones.
What is "Continued Pre-training"?
Unlike Fine-Tuning (which teaches tasks), Continued Pre-training teaches knowledge and language patterns.
Which AWS service would you use to store the Vector Index for a Knowledge Base if you want a serverless experience?
OpenSearch Serverless simplifies operations by removing the need to manage clusters/nodes.
What is the "Context Window" limit for Claude 3 Opus?
200k tokens allows you to paste entire books or codebases into the prompt.
What does "Answer Relevance" measure?
An answer can be faithful (true) but irrelevant (doesn't answer the question).
How can an Agent handle ambiguous user requests?
Good agent design includes the ability to say "I found multiple flights. Which time do you prefer?"
Quiz Progress
0 / 0 questions answered
(0%)
0 correct
Quiz Complete!
0%
📚 Study Guides
📬 Weekly DevOps, Cloud & Gen AI quizzes & guides