Gen AI (intermediate)plain Mode
Review all 15 questions for Gen AI (intermediate).Plain Q&A (No XP)
#1Multiple Choice
What is the primary benefit of Retrieval-Augmented Generation (RAG) over model fine-tuning for internal company documentation?
#2Multiple Choice
Which vector index type uses multi-layer graph structures to achieve sub-10ms Approximate Nearest Neighbor (ANN) search?
#3Multiple Choice
What prompting technique improves math and logical reasoning accuracy by forcing the LLM to output intermediate calculation steps?
#4Multiple Choice
In RAG evaluation, what does the 'Faithfulness' metric measure?
#5Multiple Choice
When should you choose Fine-Tuning instead of RAG?
#6Multiple Choice
What is the primary trade-off of using a fixed chunk size of 100 tokens versus 1000 tokens during RAG indexing?
#7Scenario
An LLM consistently fails multi-step financial calculations when given a raw prompt. Adding 'Be accurate' doesn't help. How do you restructure the prompt to fix calculation accuracy?
#8Scenario
Your vector search query for 'Error code 0x80070005' returns generic Windows troubleshooting articles instead of the specific internal post for error 0x80070005. Why did pure vector search fail?
#9Scenario
A customer support bot relies on RAG. Users report that the bot answers correctly when context is present, but invents answers when no relevant documents exist. What prompt constraint is missing?
#10Scenario
You need to evaluate 500 prompt candidate changes automatically in CI. Manually reviewing outputs is too slow. What evaluation design pattern scales this?
#11Code Fill
Complete the mathematical formula for Cosine Similarity between vector A and vector B:
similarity = (A • B) / (norm(A) * ___)#12Code Fill
Complete the RAG pipeline step that re-orders initial vector search results using a cross-encoder model:
initial_docs = vector_db.search(query, k=50)
final_docs = ___(query, initial_docs, top_k=5)#13Code Fill
Complete the keyword search algorithm acronym paired with vector search in hybrid retrieval architectures:
hybrid_results = merge_results(vector_results, ___ _results)#14Open Ended
Explain how Hybrid Search combines BM25 and Vector Search using Reciprocal Rank Fusion (RRF).
#15Open Ended