Research map / Label-efficient and robust learning / OOD and anomaly detection
GalLop: Learning global and local prompts for vision-language models
Marc Lafon, Elias Ramzi, Clément Rambour, Nicolas Audebert, Nicolas Thome
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 |
|---|---|---|---|
| combine learned global and local prompts in GalLoP (ablation) | GalLoP · ImageNet 16-shot | top-1 accuracy: 72.0 -> 75.1 | Improved |
| naively combine CoOp global and CoOp local prompts (ablation) | CoOp · ImageNet 16-shot | top-1 accuracy: 71.4 -> 69.5 | Hurt |
| combine global and local prompts in GalLoP on OOD detection (ablation) | GalLoP · ImageNet OOD benchmark | FPR95: -8.9pt FPR95 vs local only | Improved |
| keep only top-k local features (sparsity) instead of all (ablation) | local prompt learning · ImageNet 16-shot | top-1 accuracy: +18.4pt (zero-shot), +17.6pt (local prompt), +8.5pt (with linear) | Improved |
| align local prompt with a linear projection (ablation) | local prompt · ImageNet 16-shot, k=10 | top-1 accuracy: +10pt | Improved |
| add prompt dropout when learning multiple global prompts (ablation) | global prompts · ImageNet 16-shot | top-1 accuracy | Improved |
| increase local prompt scales from 1 to 4 (ablation) | local prompts · ImageNet 16-shot | top-1 accuracy: +1pt | Improved |
| increase local prompt scales beyond 4 (ablation) | local prompts · ImageNet 16-shot | top-1 accuracy | No clear effect |
| use GalLoP instead of domain-generalization-specific PromptSRC | domain generalization · ImageNet variants average | top-1 accuracy: +0.5pt average, +4.9pt ImageNet | Improved |
| use GalLoP instead of LoCoOp for OOD detection | OOD detection · ImageNet OOD benchmark | FPR95: -1.4pt FPR95, +3.6pt top-1 | Improved |
Later papers that compared with this method
2 later accepted papers report GalLoP as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- $\Delta \mathrm{Energy}$: Optimizing Energy Change During Vision-Language Alignment Improves both OOD Detection and OOD Generalization (NeurIPS 2025 · better on 1 of 2 shared results)
- Auxiliary Prompt Tuning of Vision-Language Models for Few-Shot Out-of-Distribution Detection (ICCV 2025 · better on 1 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Local-Prompt: Extensible Local Prompts for Few-Shot Out-of-Distribution Detection (ICLR 2025 · similarity 0.89)
- CoPL: Contextual Prompt Learning for Vision-Language Understanding (AAAI 2024 · similarity 0.88)
- Visual Diversity and Region-aware Prompt Learning for Zero-shot HOI Detection (NeurIPS 2025 · similarity 0.87)
- Aggregate-and-Adapt Natural Language Prompts for Downstream Generalization of CLIP (NeurIPS 2024 · similarity 0.87)
- DeCoOp: Robust Prompt Tuning with Out-of-Distribution Detection (ICML 2024 · similarity 0.87)
- PromptKD: Unsupervised Prompt Distillation for Vision-Language Models (CVPR 2024 · similarity 0.87)
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.
