Research map / Vision-language and multimodal / CLIP and zero-shot
Generalizing Vision-Language Models with Dedicated Prompt Guidance
Xinyao Li, Yinjie Min, Hongbo Chen, Zhekai Du, Fengling Li, Jingjing Li
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 GuiDG to ERM (WF) fine-tuning | ERM (WF), CLIP ViT-B/16 · ImageNet-DG 16-shot | accuracy: 66.2 -> 67.9 | Improved |
| add GuiDG to ERM (WF) in 8-shot few-shot DG | ERM (WF), CLIP ViT-B/16 · OfficeHome 8-shot | accuracy: 82.7 -> 85.2 | Improved |
| add GuiDG to ERM (WF) with full data | ERM (WF), CLIP ViT-B/16 · DomainNet full data | accuracy: 60.2 -> 61.7 | Improved |
| add GuiDG to CLIPood with full data | CLIPood, CLIP ViT-B/16 · DomainNet full data | accuracy: 63.4 -> 63.4 | No clear effect |
| add GuiDG to CLIPood on PACS | CLIPood, CLIP ViT-B/16 · PACS | accuracy: 97.3 -> 97.3 | No clear effect |
| add GuiDG to CLIPood on TerraIncognita | CLIPood, CLIP ViT-B/16 · TerraIncognita | accuracy: 60.0 -> 60.9 | Improved |
| add GuiDG to UEO (WF) on VLCS | UEO (WF), CLIP ViT-B/16 · VLCS | accuracy: 81.3 -> 84.2 | Improved |
| add GuiDG to CLIPood on single-source ImageNet 16-shot | CLIPood, CLIP ViT-B/16 · Single-source ImageNet 16-shot | accuracy: 62.7 -> 63.5 | Improved |
| add GuiDG to CLIPood on OfficeHome-Prod 16-shot | CLIPood, CLIP ViT-B/16 · OfficeHome 16-shot, Art domain | accuracy: 86.2 -> 87.6 | Improved |
| train multiple domain experts instead of a single prompt (Step 1) (ablation) | GuiDG · ImageNet-DG 16-shot ERM (WF) | avg accuracy | Improved |
| use learnable CMAttn weights instead of simple averaging (ablation) | GuiDG · ImageNet-DG 16-shot ERM (WF) | avg accuracy | Improved |
| use i.i.d. data split between Step 1 and Step 2 (ablation) | GuiDG · ImageNet-DG 16-shot ERM (WF) | avg accuracy | Improved |
Closest papers by meaning
- PracticalDG: Perturbation Distillation on Vision-Language Models for Hybrid Domain Generalization (CVPR 2024 · similarity 0.91)
- Connecting the Dots: Collaborative Fine-tuning for Black-Box Vision-Language Models (ICML 2024 · similarity 0.90)
- MMA: Multi-Modal Adapter for Vision-Language Models (CVPR 2024 · similarity 0.89)
- Adaptive Parameter Selection for Tuning Vision-Language Models (CVPR 2025 · similarity 0.89)
- Leveraging Vision-Language Models for Improving Domain Generalization in Image Classification (CVPR 2024 · similarity 0.89)
- Focusing Where Vision Matters: Selective Training for Large Vision Language Models via Visual Information Gain (ICML 2026 · similarity 0.89)
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.
