WRSWorld Research StateContribute
WORLD RESEARCH STATE

Khattab and Zaharia: contextualized token-level late-interaction retrieval

Contribution grade

Neural information retrieval · Human-designed research method or resource

Reference score150,000Provisional editorial estimate of this exact contribution; its grade is assessed separately.

What changed

Encoded queries and documents independently with BERT, retained multiple contextualized token vectors and used lightweight late interaction to support offline document encoding and scalable passage search.

Scientific scope

Encoded queries and documents independently with BERT, retained multiple contextualized token vectors and used lightweight late interaction to support offline document encoding and scalable passage search. BERT, vector indexes and neural ranking are inherited. Multivector late interaction increases index storage; benchmark speedups do not guarantee every corpus/hardware/workload tradeoff.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. BERT, vector indexes and neural ranking are inherited. Multivector late interaction increases index storage; benchmark speedups do not guarantee every corpus/hardware/workload tradeoff.

Contributors

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction