# CARE-FL: Federated Fine-Tuning of Multimodal LLMs in Resource-Constrained Edge Networks

DevFeed: [CARE-FL: Federated Fine-Tuning of Multimodal LLMs in Resource-Constrained Edge Networks](<https://devfeed.tech/articles/care-fl-federated-fine-tuning-of-multimodal-llms-in-resource-constrained-edge-networks-119107.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.06169>)

Author: Tianqi Zong, Baosheng Li, Nuocheng Yang, Binbin Chen, Marie Siew

Published: 2026-11-05T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>)

Topics: [Federated Learning](<https://devfeed.tech/topics/federated-learning.md>), [Fine-tuning](<https://devfeed.tech/topics/fine-tuning.md>), [lora](<https://devfeed.tech/topics/lora.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>)

Tags: [adaptive](<https://devfeed.tech/tags/adaptive.md>), [aggregation](<https://devfeed.tech/tags/aggregation.md>), [bandwidth](<https://devfeed.tech/tags/bandwidth.md>), [bandwidth-constraints](<https://devfeed.tech/tags/bandwidth-constraints.md>), [client](<https://devfeed.tech/tags/client.md>), [constraint](<https://devfeed.tech/tags/constraint.md>), [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [deep-q-learning](<https://devfeed.tech/tags/deep-q-learning.md>), [disaster](<https://devfeed.tech/tags/disaster.md>), [domain](<https://devfeed.tech/tags/domain.md>), [edge](<https://devfeed.tech/tags/edge.md>), [environments](<https://devfeed.tech/tags/environments.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [federated](<https://devfeed.tech/tags/federated.md>), [fine-tuning](<https://devfeed.tech/tags/fine-tuning.md>), [framework](<https://devfeed.tech/tags/framework.md>), [global](<https://devfeed.tech/tags/global.md>), [greedy](<https://devfeed.tech/tags/greedy.md>), [llms](<https://devfeed.tech/tags/llms.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [networks](<https://devfeed.tech/tags/networks.md>)

## AI overview

CARE-FL selects clients online for federated LoRA fine-tuning of multimodal LLMs under per-round latency and bandwidth limits. It estimates each client's long-term utility with Q-value decomposition and a modified Double Deep Q-Learning scheme, then uses constraint-aware ε-greedy selection to form feasible client subsets. Adaptive aggregation can give more weight to the previous model to control potentially unstable updates. In emulated heterogeneous wireless edge settings on the CrisisMMD disaster-response task, the authors report higher and more stable test accuracy than compared client-selection baselines while meeting the imposed constraints. The supplied text is an abstract and does not provide quantitative results or broader evaluation details.

## Source excerpt

arXiv:2611.06169v1 Announce Type: new Abstract: Domain-specific adaptation of Multimodal Large Language Models (MLLMs) in privacy-sensitive edge environments requires federated fine-tuning, yet client selection remains challenging due to heterogeneous multimodal data and modality availability, dynamic system conditions, and evolving client utility during training. We propose Constraint-Aware Reinforcement Learning for Efficient Federated Fine-Tuning (CARE-FL), a constrained online client-selection framework for federated MLLM Low-Rank Adaptation (LoRA) fine-tuning. CARE-FL formulates client participation as a sequential decision process and employs Q-value decomposition to address the large combinatorial joint action space through scalable client-level long-term utility estimation. These utilities are learned online using a modified Double Deep Q-Learning (DDQL) scheme, while constraint-aware $\epsilon$-greedy execution selects feasible client subsets under per-round latency and bandwidth constraints. Complementing client selection, an adaptive aggregation mechanism further controls potentially unstable global updates and places greater weight on the previous model when needed. Experiments on the CrisisMMD disaster-response task in emulated heterogeneous wireless edge environments show that CARE-FL achieves higher and more stable test accuracy than the compared client-selection baselines across different system settings, while satisfying the imposed latency and bandwidth constraints.