**Translation**
In recent years, textual embedding models, especially those built on Sentence Transformers, have emerged as key tools in the fields of information retrieval and semantic search. Multi‑vector or Late Interaction models, instead of performing early fusion of text vectors, carry out late interaction between local features of each token after extracting those features; an approach employed in models such as ColBERT and Multi‑Vector Embedding, leading to a significant improvement in ranking accuracy.
This technique enables more precise comparisons between queries and documents while preserving computational efficiency, because each token is compared individually with the target text tokens and the results are then combined. Recent research has shown that combining this approach with advanced Sentence Transformers architectures not only speeds up processing but also deepens the model’s ability to understand semantic context. Consequently, multi‑vector models with late interaction are regarded as an important advancement in information retrieval and natural language processing.
**Reasoning and Explanation**
- The original Persian passage discusses recent developments in text embedding models, emphasizing Sentence Transformers and the rise of multi‑vector (Late Interaction) approaches.
- I identified the key concepts: “Sentence Transformers,” “Late Interaction,” “ColBERT,” “Multi‑Vector Embedding,” and their impact on ranking accuracy and computational efficiency.
- I rendered each sentence into natural English, preserving the technical terminology and the logical flow of the original text.
- Paragraph breaks were kept to reflect the structure of the source, ensuring readability while maintaining fidelity to the source meaning.