AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Visual Prompt Tuning

Menglin Jia, Luming Tang, Bor-Chun Chen, Claire Cardie, Serge Belongie, Bharath Hariharan, Ser-Nam Lim

ECCV 2022 · Architectures and efficiency › Pruning and efficient tuning · approach: Efficient ViT adaptation · 2,268 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
add bias tuning to VPT-Deep (ablation)VPT-Deep · VTAB-Structuredmean accuracy: 54.98 -> 53.87Hurt
add bias tuning to VPT-Shallow (ablation)VPT-Shallow · VTAB-Naturalmean accuracy: 76.81 -> 79.78Improved
insert prompts in pixel space (Prepend-pixel) instead of latent embedding space (ablation)VPT-Shallow · VTAB-Naturalmean accuracy: drops 6.9%Hurt
concatenate a new input channel instead of prepending prompt tokens (ablation)VPT · VTAB-Naturalmean accuracy: 30 accuracy scores dropHurt
pool final prompt outputs into the classifier input (ablation)VPT-Deep · VTABaccuracy: as large as 8 pointsHurt
average-pool image patch outputs instead of CLS as final output (ablation)VPT-Deep · VTAB-Specializedaccuracy: 82.4 vs 82.3No clear effect
learn the prompt embeddings rather than keeping them frozen (ablation)VPT · VTABaccuracyImproved
share one prompt across layers (Shared-inter) instead of per-layer prompts (ablation)VPT-Deep · VTABaccuracy: 1.14x vs 1.13x paramsNo clear effect
share the same prompt among and within layers (Shared-all) (ablation)VPT-Deep · VTABaccuracyHurt
initialize prompts with class-average CLS embeddings instead of random (ablation)VPT · VTABaccuracyHurt
insert prompts in earlier layers rather than only later layers (prompt depth) (ablation)VPT-Deep · VTABaccuracyImproved
use VPT-Deep instead of full fine-tuning on ViT-BViT-B/16 · VTAB-Structuredmean accuracy: 47.64 -> 54.98Improved

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