Personalized Wireless Federated Learning for Large Language Models

Large Language Models (LLMs) have revolutionized natural language processingtasks. However, their deployment in wireless networks still face challenges,i.e., a lack of privacy and security protection mechanisms. Federated Learning(FL) has emerged as a promising approach to address these challenges. Yet, itsuffers from issues including inefficient handling with big and heterogeneousdata, resource-intensive training, and high communication overhead. To tacklethese issues, we first compare different learning stages and their features ofLLMs in wireless networks. Next, we introduce two personalized wirelessfederated fine-tuning methods with low communication overhead, i.e., (1)Personalized Federated Instruction Tuning (PFIT), which employs reinforcementlearning to fine-tune local LLMs with diverse reward models to achievepersonalization; (2) Personalized Federated Task Tuning (PFTT), which canleverage global adapters and local Low-Rank Adaptations (LoRA) tocollaboratively fine-tune local LLMs, where the local LoRAs can be applied toachieve personalization without aggregation. Finally, we perform simulations todemonstrate the effectiveness of the proposed two methods and comprehensivelydiscuss open issues.

Further reading#

• Access Paper in arXiv.org