Research map / Vision-language and multimodal / CLIP and zero-shot
Boosting Visual Reprogramming for CLIP with Dual Granularity Alignment
Jiayang Wu, Xinyang Chen, Ke Lv, Weili Guan
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add HKP (logit redistribution and aggregation) to the SG branch (ablation) | DGA · SUN, UCF, Pets, Aircraft average | accuracy (%): 75.6 -> 77.3 | Improved |
| add the SG branch (ablation) | DGA · SUN, UCF, Pets, Aircraft average | accuracy (%): 76.4 -> 77.3 | Improved |
| add UPF uncertainty-weighted crop fusion in VG branch (ablation) | DGA · SUN, UCF, Pets, Aircraft average | accuracy (%): 75.7 -> 77.3 | Improved |
| add the VG multi-scale branch (ablation) | DGA · SUN, UCF, Pets, Aircraft average | accuracy (%): 75.5 -> 77.3 | Improved |
| add VG branch (ablation) | DGA · UCF | accuracy (%): 84.2 -> 86.7 | Improved |
| add VG branch (ablation) | DGA · Aircraft | accuracy (%): 48.4 -> 51.0 | Improved |
| use DGA instead of DVP-CSE | CLIP ViT-B/16 visual reprogramming · 12 datasets, 16-shot average | accuracy (%): 79.3 -> 83.8 | Improved |
| increase number of SG layers L or VG levels E (ablation) | DGA · UCF and DTD | accuracy (%) | Improved |
| vary fusion weight lambda (ablation) | DGA · Pets, UCF, Flowers, Resisc | accuracy (%): std < 0.5% across lambda values | No clear effect |
| vary UPF entropy threshold in [0.3, 2.1] (ablation) | DGA · Pets, UCF, Flowers, Resisc | accuracy (%): fluctuations within 1.6% | No clear effect |
| apply DGA across CLIP backbones (RN50, RN101, ViT-B/32, ViT-B/16) | visual reprogramming baselines · 12-dataset average | accuracy (%) | Improved |
Closest papers by meaning
- Understanding Model Reprogramming for CLIP via Decoupling Visual Prompts (ICML 2025 · similarity 0.91)
- Bayesian-guided Label Mapping for Visual Reprogramming (NeurIPS 2024 · similarity 0.90)
- Post-pre-training for Modality Alignment in Vision-Language Foundation Models (CVPR 2025 · similarity 0.88)
- Fine-Grained Verifiers: Preference Modeling as Next-token Prediction in Vision-Language Alignment (ICLR 2025 · similarity 0.88)
- Attribute-based Visual Reprogramming for Vision-Language Models (ICLR 2025 · similarity 0.88)
- Rethinking Misalignment in Vision-Language Model Adaptation from a Causal Perspective (NeurIPS 2024 · similarity 0.88)
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.
