AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

White-Box Transformers via Sparse Rate Reduction

Yaodong Yu, Sam Buchanan, Druv Pai, Tianzhe Chu, Ziyang Wu, Shengbang Tong, Benjamin David Haeffele, Yi Ma

NeurIPS 2023 · Architectures and efficiency › Attention and transformers · approach: Token merging and masked pretraining · 22 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
scale CRATE from Tiny to LargeCRATE · ImageNet-1Ktop-1 accuracy (%): 66.7 -> 71.3Improved
use white-box CRATE-Base instead of ViT-Small at similar parameter counttransformer architecture · ImageNet-1Ktop-1 accuracy (%): 72.4 -> 70.8Hurt
CRATE-Large vs ViT-Small on CIFAR100 transfertransfer fine-tuning · CIFAR100top-1 accuracy (%): 83.2 -> 83.6No clear effect
CRATE vs ViT on Oxford-IIIT-Pets transfertransfer fine-tuning · Oxford-IIIT-Petstop-1 accuracy (%): 88.6 -> 87.4Hurt
train CRATE (learn subspaces and dictionaries) versus untrained (ablation)CRATE-Base · ImageNet-1K layer-wise analysiscompression and sparsity objective improvement across layersImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy