Guo et al.: score-aware quantization losses for inner-product search

What changed

Weighted quantization error by high-inner-product query/data pairs, optimizing approximate retrieval relevance rather than only Euclidean vector reconstruction error.

Scientific scope

Weighted quantization error by high-inner-product query/data pairs, optimizing approximate retrieval relevance rather than only Euclidean vector reconstruction error. Quantization and maximum-inner-product retrieval are inherited. Pins the original2019 six-author loss paper, not the later seven-author ScaNN systems revision or2020 implementation release. Statistical/theoretical assumptions and empirical speed–recall tradeoffs remain source-specific.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Quantization and maximum-inner-product retrieval are inherited. Pins the original2019 six-author loss paper, not the later seven-author ScaNN systems revision or2020 implementation release. Statistical/theoretical assumptions and empirical speed–recall tradeoffs remain source-specific.

Contributors

• Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.

• New Loss Functions for Fast Maximum Inner Product Search

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