AI Paper Atlas

Research map / Vision-language and multimodal / CLIP and zero-shot

Exploring Interpretability for Visual Prompt Tuning with Cross-layer Concepts

Yubin Wang, XINYANG JIANG, De Cheng, Xiangqian Zhao, Zilong Wang, Dongsheng Li, Cairong Zhao

ICLR 2026 · Vision-language and multimodal › CLIP and zero-shot · approach: Prompt learning for CLIP

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
use 4 prompted layers with decreasing prototype counts [16,13,10,7] vs 1 layer (ablation)IVPT (DinoV2-B) · CUB-200-2011consistency score: 62.5 -> 75.3Improved
extend from 4 to 5 prompted layers (ablation)IVPT (DinoV2-B) · CUB-200-2011consistency score: 75.3 -> 66.6Hurt
decreasing prototype count per layer vs constant 10 (ablation)IVPT (DinoV2-B, 4 layers) · CUB-200-2011stability score: 71.2 -> 75.9Improved
decreasing vs constant prototype count per layer (ablation)IVPT (DinoV2-B, 4 layers) · CUB-200-2011consistency score: 75.6 (constant) vs 75.3 (decreasing)No clear effect
add concept prototypes to VPT-DeepVPT-Deep (DinoV2-B) · CUB-200-2011consistency score: 14.6 -> 70.2Improved
add concept prototypes to VPT-DeepVPT-Deep (DinoV2-B) · CUB-200-2011accuracy (%): 89.1 -> 90.3Improved
full IVPT (cross-layer, fusion) vs VPT-Deep with prototypesDinoV2-B · CUB-200-2011consistency score: 70.2 -> 75.3Improved
IVPT vs Huang et al. on small backboneDinoV2-S · CUB-200-2011consistency score: 65.7 -> 63.5Hurt
add intra-region feature aggregation (ablation)IVPT baseline · CUB-200-2011accuracy (%): over 1% increaseImproved
add cross-layer prototypes and fine-to-coarse fusion (ablation)IVPT baseline · CUB-200-2011consistency and stability scoresImproved

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