AI Knowledge Information 3 — Questions and Answers
Question 1: What is the main advantage of using vector embeddings to represent knowledge compared to symbolic representations?
- They are human-readable and easily auditable
- They capture semantic similarity and enable fuzzy matching (Correct answer)
- They enforce strict logical consistency
- They reduce storage requirements to zero
Correct answer: They capture semantic similarity and enable fuzzy matching
Vector embeddings encode semantic meaning such that similar concepts cluster together in the embedding space, enabling similarity search and generalization.
Question 2: In a transformer-based language model, where is factual knowledge primarily stored?
- Attention heads exclusively
- Positional encoding layers
- Feed-forward network (FFN) layers (Correct answer)
- The tokenizer vocabulary
Correct answer: Feed-forward network (FFN) layers
Research suggests that factual associations are predominantly encoded in the feed-forward layers of transformers, which act as key-value memories.
Question 3: Which technique allows a language model to update its knowledge without full retraining?
- Prompt engineering
- Knowledge editing (model editing) (Correct answer)
- Quantization
- Beam search tuning
Correct answer: Knowledge editing (model editing)
Knowledge editing methods like ROME or MEMIT surgically modify specific weights in a trained model to update factual associations without retraining the entire model.
Question 4: What is 'entity resolution' in the context of knowledge base construction?
- Determining the resolution of images in a multimodal dataset
- Identifying when different mentions refer to the same real-world entity (Correct answer)
- Converting entity names to numerical identifiers
- Filtering low-confidence entities from the knowledge graph
Correct answer: Identifying when different mentions refer to the same real-world entity
Entity resolution (also called entity linking or deduplication) determines that different text mentions—like 'ML' and 'machine learning'—refer to the same underlying concept or entity.
Question 5: What is 'hallucination' in large language models and how does it relate to knowledge?
- The model generating adversarial examples to fool classifiers
- The model producing confident but factually incorrect statements not grounded in its training data (Correct answer)
- The model refusing to answer questions outside its knowledge cutoff
- The model generating duplicate tokens in output sequences
Correct answer: The model producing confident but factually incorrect statements not grounded in its training data
Hallucination occurs when a language model generates plausible-sounding but factually incorrect information, often because the query falls outside its reliable knowledge boundary.
Question 6: Which graph database query language is most commonly used to query property graphs in knowledge systems?
- SPARQL
- Cypher (Correct answer)
- GraphQL
- SOQL
Correct answer: Cypher
Cypher is the declarative query language used by Neo4j and other property graph databases, designed specifically for pattern matching over graph structures.
Question 7: What is the primary purpose of a 'knowledge cutoff' date in deployed AI systems?
- Limiting the number of tokens the model can process
- Indicating the date beyond which the model has no trained knowledge (Correct answer)
- Setting an expiry for cached inference results
- Defining when the model must be retrained for compliance
Correct answer: Indicating the date beyond which the model has no trained knowledge
The knowledge cutoff is the date after which training data was not included, meaning the model lacks awareness of events, discoveries, or changes that occurred after that date.
AI Knowledge Information 3
What is the main advantage of using vector embeddings to represent knowledge compared to symbolic representations?