Pith. sign in

REVIEW 2 cited by

Dual-Personalizing Adapter for Federated Foundation Models

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2403.19211 v2 pith:5F527HIR submitted 2024-03-28 cs.LG cs.AIcs.CL

Dual-Personalizing Adapter for Federated Foundation Models

classification cs.LG cs.AIcs.CL
keywords modelstest-timedistributionfederatedfoundationadapterfedfmpersonalization
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Recently, foundation models, particularly large language models (LLMs), have demonstrated an impressive ability to adapt to various tasks by fine-tuning diverse instruction data. Notably, federated foundation models (FedFM) emerge as a privacy preservation method to fine-tune models collaboratively under federated learning (FL) settings by leveraging many distributed datasets with non-IID data. To alleviate communication and computation overhead, parameter-efficient methods are introduced for efficiency, and some research adapted personalization methods to FedFM for better user preferences alignment. However, a critical gap in existing research is the neglect of test-time distribution shifts in real-world applications, and conventional methods for test-time distribution shifts in personalized FL are less effective for FedFM due to their failure to adapt to complex distribution shift scenarios and the requirement to train all parameters. To bridge this gap, we refine the setting in FedFM, termed test-time personalization, which aims to learn personalized federated foundation models on clients while effectively handling test-time distribution shifts simultaneously. To address challenges in this setting, we explore a simple yet effective solution, a Federated Dual-Personalizing Adapter (FedDPA) architecture. By co-working with a foundation model, a global adapter and a local adapter jointly tackle the test-time distribution shifts and client-specific personalization. Additionally, we introduce an instance-wise dynamic weighting mechanism that dynamically integrates the global and local adapters for each test instance during inference, facilitating effective test-time personalization. The effectiveness of the proposed method has been evaluated on benchmark datasets across different NLP tasks.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. LAARA: Layer-Aware Adaptive Rank Allocation for Parameter-Efficient Fine-Tuning

    cs.LG 2026-07 conditional novelty 5.0

    LAARA allocates LoRA ranks per layer from diagonal Fisher (gradient-based) estimates, reporting improved accuracy with fewer trainable parameters on GLUE and MathInstruct.

  2. A Survey on Efficient Federated Learning Methods for Foundation Model Training

    cs.LG 2024-01 unverdicted novelty 4.0

    A literature survey that introduces a taxonomy for computational and communication efficiency in federated learning with foundation models and discusses PEFT, framework readiness, and open research questions.