Izacard and Grave: passage-independent encoding with decoder evidence fusion

What changed

Encoded each retrieved question-passage context independently and let one decoder attend to their concatenated representations, enabling evidence fusion across more passages without joint encoder self-attention over all contexts.

Scientific scope

Encoded each retrieved question-passage context independently and let one decoder attend to their concatenated representations, enabling evidence fusion across more passages without joint encoder self-attention over all contexts. Uses inherited BM25/DPR retrieval and pretrained sequence-to-sequence models. The encoding-cost advantage does not remove decoder attention, retrieval or memory cost; more evidence need not improve every answer or prevent hallucination.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Uses inherited BM25/DPR retrieval and pretrained sequence-to-sequence models. The encoding-cost advantage does not remove decoder attention, retrieval or memory cost; more evidence need not improve every answer or prevent hallucination.

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.

• Leveraging Passage Retrieval with Generative Models for Open Domain Question Answering

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