Research map / Federated learning and privacy / Personalized and heterogeneous FL
FedPissa: Towards Federated Personalized Adaptation of Foundation Models via LoRA Subspace Mapping
Wenwen He, Wenke Huang, Yi Liu, Jian Liang, Xirui Li, Guansong Pang, Mang Ye
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 |
|---|---|---|---|
| aggregate only B matrices (keep A local) instead of aggregating both (ablation) | FedIT · DomainNet mean (feature non-IID) | accuracy (%): 43.73 -> 51.68 | Improved |
| add decorrelated subspace projection (PMA) on top of B aggregation (ablation) | FedPissa · DomainNet mean (feature non-IID) | accuracy (%): 51.68 -> 51.80 | Improved |
| set decorrelation strength lambda to 0.1 rather than 0.5 (ablation) | FedPissa · GLUE mean (feature non-IID) | accuracy (%): 76.85 -> 83.88 | Improved |
| use smaller lambda 0.05 rather than 0.1 on vision (ablation) | FedPissa · DomainNet mean (feature non-IID) | accuracy (%): 50.55 -> 51.80 | Improved |
| use tail rank 7 versus 5 in decorrelation subspace (ablation) | FedPissa · GLUE (feature non-IID) | mean accuracy (%): Rank5 83.88 vs Rank7 84.33 mean | Mixed |
| use FedPissa instead of FedIT | RoBERTa-large federated LoRA · GLUE mean (feature and label non-IID, beta=0.1) | accuracy (%): 62.46 -> 89.09 | Improved |
| use FedPissa instead of single-LoRA FedSA | RoBERTa-large federated LoRA · GLUE mean (feature non-IID) | accuracy (%): 82.05 -> 83.88 | Improved |
| use FedPissa instead of FFA-LoRA on SST-2 | RoBERTa-large federated LoRA · SST-2 (feature non-IID) | accuracy (%): 94.70 -> 91.40 | Hurt |
| increase LoRA rank from 8 to 16 (ablation) | FedPissa · GLUE tasks (feature non-IID) | accuracy (%) | No clear effect |
| vary local epochs (2 to 5) and global rounds (25 to 35) (ablation) | FedPissa · GLUE SST-2/QNLI/MRPC mean | accuracy (%): 87.57 (2 epochs), 91.00 (4 epochs) | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FedSA — FedSA: A Unified Representation Learning via Semantic Anchors for Prototype-based Federated Learning (AAAI 2025) · better on 4 of 4 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
- Adaptive LoRA Experts Allocation and Selection for Federated Fine-Tuning (NeurIPS 2025 · similarity 0.89)
- Selective Aggregation for Low-Rank Adaptation in Federated Learning (ICLR 2025 · similarity 0.89)
- Robust Federated Finetuning of LLMs via Alternating Optimization of LoRA (NeurIPS 2025 · similarity 0.88)
- Ravan: Multi-Head Low-Rank Adaptation for Federated Fine-Tuning (NeurIPS 2025 · similarity 0.87)
- FedSVD: Adaptive Orthogonalization for Private Federated Learning with LoRA (NeurIPS 2025 · similarity 0.87)
- Fed-SB: A Silver Bullet for Extreme Communication Efficiency and Performance in (Private) Federated LoRA Fine-Tuning (NeurIPS 2026 · 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.
