Research map / Vision-language and multimodal / CLIP and zero-shot
A Retrospect to Multi-prompt Learning across Vision and Language
Ziliang Chen, Xin Huang, Quanlong Guan, Liang Lin, Weiqi Luo
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 |
|---|---|---|---|
| generate prompts with EMPL energy-based sampler for CoOp | CoOp · Base-to-new generalization (avg) | New-class accuracy: 63.22 -> 70.93 | Improved |
| generate prompts with EMPL energy-based sampler for ProDA | ProDA · Base-to-new generalization (avg) | Harmonic mean: 76.65 -> 77.39 | Improved |
| add EMPL to CoOp/ProDA per dataset in base-to-new | CoOp and ProDA · 10 datasets base-to-new | Base accuracy per dataset: 7 of 10 (CoOp), 6 of 10 (ProDA) improved | Mixed |
| add EMPL to CoOp and ProDA in cross-domain generalization | CoOp and ProDA · ImageNet variants | accuracy: improves 7 of 8 target settings; CoOp ImageNet 71.51 -> 70.89 | Mixed |
| add EMPL to CoOp in cross-dataset transfer | CoOp · Cross-dataset transfer | Avg target accuracy: 63.88 -> 66.49 | Improved |
| add EMPL to CoOp for few-shot image-text retrieval | CoOp · MSCOCO retrieval (1% data) | R@1: 53.58 -> 56.85 (+3.37) | Improved |
| apply SGLD sampling through more (low-level) transformer layers (ablation) | EMPL (+CoOp) · DTD base-to-new | Harmonic mean: 63.32 -> 61.62 (2 vs 9 layers); 0.048 -> 0.278 sec/iter | No clear effect |
| increase lambda weight on open-vocabulary energy term (ablation) | EMPL (+CoOp) · DTD base-to-new | new-class accuracy | Mixed |
| draw more prompt samples from the energy-based distribution (ablation) | EMPL (+CoOp) · DTD base-to-new | Harmonic mean | Improved |
| add more learnable prompts (CLIP to CoOp to ProDA) to shrink modality gap | prompt learning · MsCOCO modality gap | gap magnitude | Improved |
Closest papers by meaning
- CoPL: Contextual Prompt Learning for Vision-Language Understanding (AAAI 2024 · similarity 0.92)
- Weighted Multi-Prompt Learning with Description-free Large Language Model Distillation (ICLR 2025 · similarity 0.90)
- LaViP: Language-Grounded Visual Prompting (AAAI 2024 · similarity 0.90)
- COMMA: Co-articulated Multi-Modal Learning (AAAI 2024 · similarity 0.90)
- IPO: Interpretable Prompt Optimization for Vision-Language Models (NeurIPS 2024 · similarity 0.90)
- PromptKD: Unsupervised Prompt Distillation for Vision-Language Models (CVPR 2024 · similarity 0.90)
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.
