Research map / Federated learning and privacy / Personalized and heterogeneous FL
Federated Prompt-Tuning with Heterogeneous and Incomplete Multimodal Client Data
Thu Hang Phung, Duong M. Nguyen, Thanh Trung Huynh, Quoc Viet Hung Nguyen, Trong Nghia Hoang, Phi Le Nguyen
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 inter-client prompt set (FED-PRIME vs FED-INTRA only) (ablation) | FED-PRIME · UPMC Food-101 train Miss Text, test ~Train | accuracy: 61.71 -> 78.88 | Improved |
| add intra-client prompt set (FED-PRIME vs FED-INTER only) (ablation) | FED-PRIME · UPMC Food-101 train Miss Text, test ~Train | accuracy: 54.82 -> 78.88 | Improved |
| add inter-client prompt set (ablation) | FED-PRIME · UPMC Food-101 train Miss Image, test ~Train | accuracy: 22.84 -> 90.55 | Improved |
| add intra-client prompt set (ablation) | FED-PRIME · MM-IMDB train Miss Both, test ~Train | F1-macro: 26.53 -> 32.01 | Improved |
| add inter-client prompt set (ablation) | FED-PRIME · MM-IMDB train Miss Both, test ~Train | F1-macro: 11.90 -> 32.01 | Improved |
| increase prompt pool size from 10 to 20 (ablation) | FED-PRIME · UPMC Food-101 | top-1 accuracy: 73.94 -> 82.61 | Improved |
| increase prompt pool size from 20 to 30 (ablation) | FED-PRIME · UPMC Food-101 | top-1 accuracy: 82.61 -> 83.21 | No clear effect |
| use FED-PRIME instead of best baseline FedAvg-style prompt tuning | FEDAVG-P · UPMC Food-101 train Miss Both, test ~Train | accuracy: 56.32 -> 84.44 | Improved |
| use FED-PRIME instead of FEDMSPLIT-P on Miss Image training | FEDMSPLIT-P · UPMC Food-101 train Miss Image, test ~Train | accuracy: 74.16 -> 90.55 | Improved |
| prepend prompts only to first MSA layer instead of first six layers | FEDAVG-P · UPMC Food-101 | GPU memory (GB): 31.70 -> 16.80 | Improved |
| test with image only after Miss Image training | FED-PRIME · UPMC Food-101 | accuracy: 90.55 -> 54.14 | Hurt |
Closest papers by meaning
- Probabilistic Federated Prompt-Tuning with Non-IID and Imbalanced Data (NeurIPS 2024 · similarity 0.89)
- Unlocking the Potential of Prompt-Tuning in Bridging Generalized and Personalized Federated Learning (CVPR 2024 · similarity 0.89)
- Federated Adaptive Prompt Tuning for Multi-Domain Collaborative Learning (AAAI 2024 · similarity 0.88)
- FedPHA: Federated Prompt Learning for Heterogeneous Client Adaptation (ICML 2025 · similarity 0.88)
- Learning Reconfigurable Representations for Multimodal Federated Learning with Missing Data (NeurIPS 2025 · similarity 0.88)
- DiPrompT: Disentangled Prompt Tuning for Multiple Latent Domain Generalization in Federated Learning (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.
