ML//RAG//embedding model

- Dedicated models trained to produce dense vector representations optimized for retrieval.

Dedicated models trained to produce dense vector representations optimized for retrieval.

SentenceTransformers, OpenAI embeddings, Nomic Embed, Jina v3, each tuned for semantic similarity.

Matryoshka embeddings: train once, truncate vectors to any dimension: flexible quality/cost tradeoff.

Evaluated on MTEB leaderboard (known overfitting, but still the de-facto benchmark)