RAG

Question 1

What is Retrieval-Augmented Generation (RAG)?

Answer 1

An approach that enhances LLMs by retrieving and integrating external knowledge sources at query time.

Question 2

How does Document-based RAG differ from Traditional LLMs?

Answer 2

Document-based RAG adds factual grounding by retrieving relevant text passages before generation.

Question 3

What additional benefit does Knowledge Graph RAG provide over Document-based RAG?

Answer 3

It adds structured precision through retrieval of triples and multi-hop graph relations.

Question 4

What is Hybrid Multi-source RAG?

Answer 4

A system combining multiple retrieval sources (text, graph, embeddings) for comprehensive knowledge integration.

Question 5

List the four core KG-RAG components outlined in the lecture.

Answer 5

1) Query understanding & translation
2) Knowledge graph retrieval
3) Context formatting & integration
4) Enhanced response generation

Question 6

Name three application domains for KG-RAG mentioned in the lecture.

Answer 6

Question answering, domain-specific assistance (e.g., medical, research), enterprise search.

Question 7

What are the three initial challenges highlighted for KG-RAG?

Answer 7

Query translation, schema alignment, handling incomplete knowledge.

Question 8

Define the ‘query translation accuracy’ challenge in KG-RAG.

Answer 8

Converting complex natural-language queries into precise SPARQL while handling ambiguity.

Question 9

What is the ‘schema alignment’ challenge in KG-RAG?

Answer 9

Matching user terminology with the knowledge graph schema and dealing with different formats.

Question 10

Explain the ‘incomplete knowledge’ challenge in KG-RAG.

Answer 10

Handling queries that require missing information by combining KG retrieval with LLM generation.

Question 11

What does the ‘context integration’ challenge involve?

Answer 11

Balancing level of detail versus conciseness and formatting graph data for effective LLM understanding.

Question 12

What is meant by the ‘reasoning quality’ challenge?

Answer 12

Ensuring the LLM correctly utilizes retrieved KG facts and avoids inference errors despite correct data.

Question 13

Describe the ‘performance optimization’ challenge in KG-RAG.

Answer 13

Reducing latency in multi-step processes and scaling solutions to very large knowledge graphs.