BWhat changed
Evolved a learned memory-management network conditioned on attention values to select retained KV context separately across layers/heads, studying transfer across model/task settings.
Scientific scope
Evolved a learned memory-management network conditioned on attention values to select retained KV context separately across layers/heads, studying transfer across model/task settings. KV eviction, attention features and evolution have predecessors. Universally applicable interface does not mean universal performance preservation, lossless recall or free attention computation. Original October2024 version is distinct from December expanded experiments and later ChouBun release.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. KV eviction, attention features and evolution have predecessors. Universally applicable interface does not mean universal performance preservation, lossless recall or free attention computation. Original October2024 version is distinct from December expanded experiments and later ChouBun release.
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.
- An Evolved Universal Transformer Memory
- Cetin et al.: evolved attention-conditioned KV memory pruning — author/institutional project
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