BWhat changed
Jointly trained a prompt-conditioned sparsity predictor and language model to choose structured feed-forward parameter masks, keeping selected parameters fixed for an input or task to reduce repeated weight loading.
Scientific scope
Jointly trained a prompt-conditioned sparsity predictor and language model to choose structured feed-forward parameter masks, keeping selected parameters fixed for an input or task to reduce repeated weight loading. Structured pruning, conditional computation and differentiable top-k have predecessors. Fixed masks trade adaptation granularity for loading efficiency. Original2025 method is distinct from its later flash-resident AFM3 product-system integration; benchmarks do not guarantee equal quality for every task.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Structured pruning, conditional computation and differentiable top-k have predecessors. Fixed masks trade adaptation granularity for loading efficiency. Original2025 method is distinct from its later flash-resident AFM3 product-system integration; benchmarks do not guarantee equal quality for every task.
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.
- Instruction-Following Pruning for Large Language Models
- Hou et al.: instruction-conditioned structured parameter pruning — official institutional account
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