AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Token Compensator: Altering Inference Cost of Vision Transformer without Re-Tuning

Shibo Jie, Yehui Tang, Jianyuan Guo, Zhi-Hong Deng, Kai Han, Yunhe Wang

ECCV 2024 · Architectures and efficiency › Attention and transformers · approach: Hybrid and interpretable ViT designs · 6 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
plug ToCom into models trained without compression, target r=16ToMe on DeiT-B · VTAB-1k Specializedavg accuracy: 80.0 -> 83.6Improved
plug ToCom into models trained without compression, target r=16ToMe on DeiT-B · VTAB-1k Naturalavg accuracy: 73.7 -> 75.8Improved
plug ToCom into models trained without compression, target r=16ToMe on DeiT-B · VTAB-1k Structuredavg accuracy: 56.7 -> 57.9Improved
add ToCom with source r=0 on CIFAR100 and FGVC, target r=8 and 16ToMe on DeiT-B · CIFAR100 and FGVC averageaccuracy: +0.5 (r=8), +1.7 (r=16)Improved
add ToCom for training-time acceleration (target r lower than source r)ToMe on DeiT-B · CIFAR100, FGVC, VTAB-1kaccuracyMixed
add ToCom to ToMe for semantic segmentation at target r=16ToMe on DeiT-B384 + Segmenter · ADE20k valmIoU (SS): 41.3 -> 43.4Improved
use a single shared LoRA (h=32) instead of 16 groups of adjacent-degree LoRA (ablation)ToCom · VTAB-1kavg accuracyHurt
replace self-distillation loss with classification cross-entropy loss (ablation)ToCom · VTAB-1k (source r > target r)avg accuracyHurt
use separate LoRA for inverse direction instead of subtracting (invertibility assumption) (ablation)ToCom · VTAB-1kavg accuracyHurt
increase LoRA hidden dimension h beyond 2 (ablation)ToCom · VTAB-1kavg accuracyNo clear effect
apply ToCom to EViT token pruning instead of ToMeEViT · VTAB-1kavg accuracyImproved

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