BWhat changed
Introduced attention-based distillation through a dedicated token and a data-efficient supervised ViT training recipe; the later III study simplified and strengthened supervised augmentation/training baselines.
Scientific scope
Introduced attention-based distillation through a dedicated token and a data-efficient supervised ViT training recipe; the later III study simplified and strengthened supervised augmentation/training baselines. Inherits ViT, augmentation and knowledge distillation; III inherits ResNet-style recipes rather than a new transformer architecture. Teacher/label/training costs remain; data-efficient does not imply absence of external prior knowledge. Original benchmarks are not universal current rankings.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Inherits ViT, augmentation and knowledge distillation; III inherits ResNet-style recipes rather than a new transformer architecture. Teacher/label/training costs remain; data-efficient does not imply absence of external prior knowledge. Original benchmarks are not universal current rankings.
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.
- Training data-efficient image transformers & distillation through attention
- DeiT III: Revenge of the ViT
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