English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Multi-vector retrieval methods combine the merits of sparse (e.g. BM25) anddense (e.g. DPR) retrievers and have achieved state-of-the-art performance onvarious retrieval tasks. These methods, however, are orders of magnitude slowerand need much more space to store their indices compared to their single-vectorcounterparts. In this paper, we unify different multi-vector retrieval modelsfrom a token routing viewpoint and propose conditional token interaction viadynamic lexical routing, namely CITADEL, for efficient and effectivemulti-vector retrieval. CITADEL learns to route different token vectors to thepredicted lexical ``keys'' such that a query token vector only interacts withdocument token vectors routed to the same key. This design significantlyreduces the computation cost while maintaining high accuracy. Notably, CITADELachieves the same or slightly better performance than the previous state of theart, ColBERT-v2, on both in-domain (MS MARCO) and out-of-domain (BEIR)evaluations, while being nearly 40 times faster. Code and data are available athttps://github.com/facebookresearch/dpr-scale.

CITADEL: Conditional Token Interaction via Dynamic Lexical Routing for Efficient and Effective Multi-Vector Retrieval