Fine-Tuning a 3B-Parameter LLM on a Smartphone: Characterizing Sustained Training

arXiv:2610.06325 · cs.DC, cs.AI, cs.PF · Submitted 2026-10-05 · Read on arXiv

cs.DC, cs.AI, cs.PF

Submitted: 2026-10-05

Updated: 2026-10-05

Comments: 15 pages, 7 figures, 8 tables. Code and data: https://github.com/gordofreemo/mobile_LoRA_ft

Code: https://github.com/gordofreemo/mobile_LoRA_ft

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Fine-Tuning — 1× in this paper · explained in 5 episodes

Sources

• Federated Fine-Tuning of Sparsely-Activated Large Language Models on Resource-Constrained Devices
• CE-LoRA: Computation-Efficient LoRA Fine-Tuning for Language Models
• Training Deep Nets with Sublinear Memory Cost
• QLoRA: Efficient Finetuning of Quantized LLMs
• GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers
• Gemini: A Family of Highly Capable Multimodal Models
• EdgeFlowerTune: Evaluating Federated LLM Fine-Tuning Under Realistic Edge System Constraints
• MobileFineTuner: A Mobile-Native Framework for On-Device LLM Fine-Tuning in Real-World Embedded AI Applications
• LoRA: Low-Rank Adaptation of Large Language Models
• Apple Intelligence Foundation Language Models: Tech Report 2025
• MobiLLM: Enabling LLM Fine-Tuning on the Mobile Device via Server Assisted Side Tuning
• AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration
• ExecuTorch -- A Unified PyTorch Solution to Run AI Models On-Device
• Energy-Efficient GPU DVFS for Fine-Tuning of SLMs on Resource-constrained Embedded Devices
• Qwen3 Technical Report
• LST: Ladder Side-Tuning for Parameter and Memory Efficient Transfer Learning
• Parameter Efficiency Is Not Memory Efficiency: Rethinking Fine-Tuning for On-Device LLM Adaptation
• LLM Inference at the Edge: Mobile, NPU, and GPU Performance Efficiency Trade-offs Under Sustained Load
• Split learning for health: Distributed deep learning without sharing raw patient data
• PELM: Power Efficient On-Device LLM Inference with Speculative Decoding and Dynamic Voltage Frequency Scaling

Related papers

• iScheduler: Reinforcement Learning-Driven Continual Optimization for Large-Scale Resource Investment Problems
• SAMM: Sharded Automated Market Maker
• InferScale: GPU-Native KV Injection for Personalized LLM Serving
• Vigil: Accountable Liveness against Selective Silence
• Steelhead: Interleaving Partially Synchronous and Asynchronous Commit Rules on a Shared DAG
• Democratizing MoE inference on commodity GPUs with CoMoE