問題文
A team loaded embeddings from a 384 dimension model and embeddings from a 768 dimension model into the same flexible VECTOR column and now needs one similarity search across the whole table. Which action lets that search return meaningful neighbors?
選択肢
- Regenerate every row with one embedding model so that all stored vectors share one semantic space.
- Convert the shorter vectors to the wider format so that all stored vectors share one dimension count.
- Choose a distance metric that tolerates a different dimension count on each side.
- Declare the column with a fixed dimension count so that the database rescales the vectors.