Multi-vector models
AssessTechniques
Representation models that use multiple vectors per item to improve relevance matching.
Why it's here
Placed in Assess: 1 article(s) of evidence from 1 source(s), led by research-stage coverage, with 1 in the last 30 days. Confidence 24%. Low accumulated evidence, so it defaults conservatively pending more signal.
Evidence (1)
- 6InfoQ·9/16/2026researchPinterest Upgrades Manas with Quantization for More Efficient Search
Pinterest Engineering has improved its Manas search platform to handle large-scale data more efficiently for search and discovery. The system uses Scalar and Product Quantization to reduce memory use while preserving high recall, and it is moving toward multi-vector models with SSD-backed storage for better relevance matching.