How do you handle embedding model updates without breaking retrieval?

hard

Answer

$New embedding model = new vector space.$
$Options: (1) Re-embed entire corpus + swap.$
$Time-consuming for large data.$
$(2) Dual-serving: keep old + new indexes in parallel; migrate progressively.$
$(3) A/B test to validate quality before full switch.$
$(4) Query-side model translation (rare).$
$Pitfall: use same model for query + index — asymmetric = disaster.$
$Track model version per index.$
$Modern: Matryoshka embeddings allow using different dim of same model = smoother migration.$

Check yourself — multiple choice

Just swap
New model = new vector space; options: re-embed corpus (slow) / dual-index parallel migrate / A/B validate / never asymmetric query vs index (disaster); track version; Matryoshka = smoother dim migration
Random
Not real

Embedding update: re-embed / dual-index / A/B; never asymmetric.

#llmops

215 interview questions in this topic.