Research map / Vision-language and multimodal / CLIP and zero-shot
FedMVP: Federated Multimodal Visual Prompt Tuning for Vision-Language Models
Mainak Singha, Subhankar Roy, Sarthak Mehrotra, Ankit Jha, Moloud Abdar, Biplab Banerjee, Elisa Ricci
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 PromptFormer cross-attention prompts to CLIP (ablation) | ZS-CLIP · Base-to-new (9 datasets avg) | harmonic mean: 74.24 -> 75.94 | Improved |
| add consistency loss L_con (ablation) | PromptFormer · Base-to-new (9 datasets avg) | harmonic mean: 75.94 -> 76.27 | Improved |
| add LoRA fine-tuning of PromptFormer (ablation) | FedMVP · Base-to-new (9 datasets avg) | harmonic mean: 77.41 -> 78.27 | Improved |
| add PromptFormer cross-attention prompts to CLIP (ablation) | ZS-CLIP · DomainBed MSST DG (avg 5 datasets) | accuracy: 70.41 -> 71.85 | Improved |
| add consistency loss L_con (ablation) | PromptFormer · DomainBed MSST DG (avg 5 datasets) | accuracy: 71.85 -> 72.14 | Improved |
| add LoRA fine-tuning of PromptFormer (ablation) | FedMVP · DomainBed MSST DG (avg 5 datasets) | accuracy: 72.58 -> 73.02 | Improved |
| use visual prompting instead of textual prompting (ablation) | FedMVP · Base-to-new (9 datasets avg) | harmonic mean: 73.56 -> 78.27 | Improved |
| use visual prompting instead of multimodal prompting (ablation) | FedMVP · DomainBed MSST DG (avg 5 datasets) | accuracy: 69.14 -> 73.02 | Improved |
| use visual prompting instead of textual prompting (ablation) | FedMVP · DomainBed SSMT DG (avg 5 datasets) | accuracy: 65.89 -> 72.63 | Improved |
| use LLM attributes via PromptFormer cross-attention instead of DESC scoring alone | FedMVP · DomainBed MSST DG (avg 5 datasets) | accuracy: 69.89 -> 72.24 | Improved |
| use LLM attributes via PromptFormer cross-attention instead of DESC scoring alone | FedMVP · Base-to-new (9 datasets avg) | harmonic mean: 75.18 -> 77.52 | Improved |
| compare FedMVP against ZS-CLIP on OxfordPets and StanfordCars cross-dataset | FedMVP · Cross-dataset generalization | accuracy: -1.33 on OxfordPets, -4.56 on EuroSAT vs best baseline | Hurt |
Later papers that compared with this method
1 later accepted paper reports FedMVP as a baseline in their results tables; 0 of them report a better number on most of the results they share with it.
- Prompt-Robust Vision-Language Models via Meta-Finetuning (ICLR 2026 · better on 1 of 2 shared results)
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
- Federated Disentangled Tuning with Textual Prior Decoupling and Visual Dynamic Adaptation (ICML 2025 · similarity 0.91)
- Learning to Learn Better Visual Prompts (AAAI 2024 · similarity 0.91)
- SA²VP: Spatially Aligned-and-Adapted Visual Prompt (AAAI 2024 · similarity 0.91)
- IPO: Interpretable Prompt Optimization for Vision-Language Models (NeurIPS 2024 · similarity 0.90)
- Global and Local Prompts Cooperation via Optimal Transport for Federated Learning (CVPR 2024 · similarity 0.90)
- Mixture of Experts Made Personalized: Federated Prompt Learning for Vision-Language Models (ICLR 2025 · 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.
