Research map / Federated learning and privacy / Personalized and heterogeneous FL
Enhancing Visual Representation with Textual Semantics: Textual Semantics-Powered Prototypes for Heterogeneous Federated Learning
Xinghao Wu, Jianwei Niu, Xuefeng Liu, Guogang Zhu, Jiayuan Zhang, Shaojie Tang, Wei Chen
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 handcrafted-prompt text prototypes from a PLM (ablation) | FedTSP-CLIP · CIFAR-100 (Dir 0.5) | test accuracy (%): 23.47 -> 26.41 | Improved |
| add LLM-generated class descriptions as prompts (ablation) | FedTSP-BERT · CIFAR-100 (Dir 0.5) | test accuracy (%): 24.63 -> 25.76 | Improved |
| add trainable prompts to align text and image modality (ablation) | FedTSP-BERT · CIFAR-100 (Dir 0.5) | test accuracy (%): 24.63 -> 26.53 | Improved |
| add trainable prompts to align text and image modality (ablation) | FedTSP-CLIP · CIFAR-100 (Dir 0.5) | test accuracy (%): 26.41 -> 26.54 | No clear effect |
| combine LLM descriptions with trainable prompts (ablation) | FedTSP-BERT · CIFAR-100 (Dir 0.5) | test accuracy (%): 26.53 -> 27.01 | Improved |
| richer prompt text (Manual, Short, Long LLM descriptions) (ablation) | FedTSP · CIFAR-100 | test accuracy (%) | Improved |
| swap the LLM used for descriptions (GPT-4o, Claude-4-sonnet, Gemini-2.5-pro, LLaMA3-8B) (ablation) | FedTSP · CIFAR-100 | test accuracy (%) | No clear effect |
| add local DP noise to uploaded class embeddings (epsilon 1) (ablation) | FedTSP-CLIP · CIFAR-100 (Dir 0.5) | test accuracy (%): 26.92 -> 26.34 | No clear effect |
| replace server-learned or fixed prototypes with PLM text prototypes | FedProto/FedTGP/AlignFed · CIFAR-10/100, Tiny ImageNet (HtFE9) | test accuracy (%): up to 4.20% | Improved |
| use FedTSP in model-homogeneous GFL | FedAvg-style global model · CIFAR-100 (GFL, Dir 1.0) | test accuracy (%): 34.38 vs FedFA 34.69 | Mixed |
| fine-tune only classifier vs full fine-tune (ablation) | PFL variants · CIFAR-100 PFL | test accuracy (%) | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- FedFA — FedFA: Federated Feature Augmentation (ICLR 2023) · better on 0 of 1 shared result
- FedBABU — FedBABU: Toward Enhanced Representation for Federated Image Classification (ICLR 2022) · 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
- FedSA: A Unified Representation Learning via Semantic Anchors for Prototype-based Federated Learning (AAAI 2025 · similarity 0.91)
- Beyond Federated Prototype Learning: Learnable Semantic Anchors with Hyperspherical Contrast for Domain-Skewed Data (AAAI 2025 · similarity 0.90)
- Taming Cross-Domain Representation Variance in Federated Prototype Learning with Heterogeneous Data Domains (NeurIPS 2024 · similarity 0.89)
- FedTGP: Trainable Global Prototypes with Adaptive-Margin-Enhanced Contrastive Learning for Data and Model Heterogeneity in Federated Learning (AAAI 2024 · similarity 0.87)
- FedDAT: An Approach for Foundation Model Finetuning in Multi-Modal Heterogeneous Federated Learning (AAAI 2024 · similarity 0.87)
- Federated Model Heterogeneous Matryoshka Representation Learning (NeurIPS 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.
