Quiz: Building Knowledge Graphs from Text with LLMs, Safely

Quiz Summary

0 of 3 Questions completed

Questions:

You have already completed the quiz before. Hence you can not start it again.

Quiz is loading…

You must sign in or sign up to start the quiz.

You must first complete the following:

Results

Quiz complete. Results are being recorded.

Results

0 of 3 Questions answered correctly

Your time:

You have reached 0 of 0 point(s), (0)

Question 1 of 3

1. Question

The lesson breaks every LLM-to-KG pipeline into five stages and says mainstream frameworks hand you only the first two. Which stages are left for you to build?

Chunking and extraction, since frameworks supply only resolution, normalisation and validation
Correct Incorrect Correct answer

Resolution, normalisation and validation, which is where the lesson says the real work and most production failures live
Correct Incorrect Correct answer

Only validation, since LangChain and LlamaIndex already resolve and normalise entities
Correct Incorrect Correct answer

Only chunking, since every other stage is covered by neo4j-graphrag
Correct Incorrect Correct answer

Correct / 1 Points

Incorrect / 1 Points
Question 2 of 3

2. Question

What numbers does Apple’s ODKE+ (arXiv 2509.04696) report, and what produced them?

55 percent higher fact recall and 40 percent higher answer correctness across four LLMs, from grounding retrieval in a curated domain ontology
Correct Incorrect Correct answer

A 61 percent reduction in hallucinations from SPARQL-mediated answering over an RDF knowledge graph
Correct Incorrect Correct answer

92 percent semantic alignment between induced and human-crafted schemas at a scale of 50 million documents
Correct Incorrect Correct answer

A 35 percent fall in hallucinated extractions and precision rising from 91 to 98.8 percent, from grounding open-domain extraction in ontology snippets with an LLM judge and a corroborator
Correct Incorrect Correct answer

Correct / 1 Points

Incorrect / 1 Points
Question 3 of 3

3. Question

What is the lesson’s rule about merge decisions in the entity resolution stage?

A wrong merge is worse than a missed one, because it manufactures relations between things that were never related and no retrieval layer can undo it later, so every merge must be logged and reversible
Correct Incorrect Correct answer

A missed merge is worse than a wrong one, because lost recall is invisible whereas a bad merge shows up immediately in queries
Correct Incorrect Correct answer

Both errors are equally recoverable downstream, so resolution can safely be deferred until after validation
Correct Incorrect Correct answer

Merges should always be made automatically without review, since LLM adjudication is more consistent than human curation
Correct Incorrect Correct answer

Correct / 1 Points

Incorrect / 1 Points

Quiz: Building Knowledge Graphs from Text with LLMs, Safely

Quiz Summary

Information

Results

Results

Categories

1. Question

2. Question

3. Question