New framework enables robust LLM fine-tuning on edge devices considering thermal constraints

Researchers have developed Thermo-FL, a novel framework for federated fine-tuning of large language models on edge devices. This approach addresses challenges posed by hardware instability and adversarial attacks by incorporating thermal awareness into the training process. Thermo-FL dynamically adjusts local adapter training and update transmission density based on device temperature, while a robust server-side aggregation pipeline, TERRA, filters and validates sparse updates to maintain model integrity and performance on benchmarks like BoolQ and GSM8K. AI

IMPACT Enhances the feasibility of deploying and adapting large language models on resource-constrained edge devices by addressing thermal and security challenges.

RANK_REASON The cluster describes a new research paper detailing a novel framework for federated fine-tuning of LLMs. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• BoolQ
• Edge artificial intelligence
• GSM8K
• Jetson
• large language models
• LoRA+

• paper
• infra
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →