AWhat changed
Related linear attention to normalized outer-product fast-weight programmers and replaced additive writes with a learned-strength delta update that corrects the value currently associated with a key.
Scientific scope
Related linear attention to normalized outer-product fast-weight programmers and replaced additive writes with a learned-strength delta update that corrects the value currently associated with a key. Fast weights date to the1990s and error-correcting delta rules are older. Acquired contribution is their differentiable linear-attention formulation, not invention of memory/delta learning. Later DeltaNet sequence parallelization is distinct; no unlimited-capacity claim.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Fast weights date to the1990s and error-correcting delta rules are older. Acquired contribution is their differentiable linear-attention formulation, not invention of memory/delta learning. Later DeltaNet sequence parallelization is distinct; no unlimited-capacity claim.
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.
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