Embeddings Quiz

Answers and explanations appear as you take the quiz.

1. What is an embedding in the numerical representation sense used here?
   • A dictionary definition that must be stored verbatim
   • A required compression format for image files
   • A vector representing an item for a model or comparison task
   • An exact database row identifier only
2. What can closeness in a trained semantic embedding space indicate?
   • Similarity learned under that model’s objective
   • A guarantee that two statements are equally true
   • An identical byte sequence in every case
   • A universal relation independent of training
3. Which metric is commonly used to compare two embeddings?
   • File size difference
   • Alphabetical ordering
   • Word count
   • Cosine similarity
4. What tradeoff can an approximate embedding index such as HNSW introduce?
   • A guarantee of complete exact retrieval in every configuration
   • Faster search with possible missed exact nearest neighbors
   • A required change of every vector to a word definition
   • Automatic removal of all training-data bias
5. What does the classic 'king - man + woman ≈ queen' example demonstrate about embeddings?
   • They can capture relationships through vector arithmetic
   • They store exact definitions of words
   • They only work for royalty terms
   • They are random and meaningless
6. Which is a real risk when using embeddings?
   • They cannot be stored in a database
   • They only work in English
   • They can inherit and reflect biases present in their training data
   • They make all searches exact-match only
7. Two unit-length embedding vectors have dot product 0.8. What is their squared Euclidean distance?
   • 0.2
   • 0.4
   • 0.8
   • 1.6
8. When does an inner-product score equal cosine similarity for two nonzero embedding vectors?
   • Whenever their dimensions are equal
   • Whenever they came from two unrelated model families
   • Only when their coordinates are all zero
   • When both vectors have been normalized to unit length