🤖 AI Summary
Deploying personalized large language models on resource-constrained edge devices faces significant challenges in GPU memory consumption, energy usage, and computational latency. This work systematically evaluates five parameter-efficient fine-tuning (PEFT) methods—LoRA, LoRA+, QLoRA, BitFit, and full fine-tuning—on small language models (SLMs) based on both Transformer and Mamba architectures across general and personalized tasks. The study introduces NetScore-E/M, a unified evaluation framework that jointly accounts for energy and memory constraints, and establishes an energy-aware PEFT selection strategy. Experimental results demonstrate that LoRA+ achieves the best overall performance, while QLoRA substantially reduces memory footprint. Among evaluated models, TinyLlama-1.1B consistently excels across multiple benchmarks, confirming the feasibility of deploying SLMs with PEFT for efficient on-device personalization.
📝 Abstract
Despite rapid advances in large language models (LLMs), deploying and personalizing them on resource-constrained devices remains impractical due to high VRAM, time, and energy costs. Parameter-Efficient Fine-Tuning (PEFT) of Small Language Models (SLMs) offers a promising alternative, yet few studies compare PEFT methods across architectures using both general and personalization benchmarks while accounting for energy consumption. We compare five fine-tuning approaches (Full Fine-Tuning, LoRA, LoRA+, QLoRA, and BitFit) on four SLMs from two families (Transformer-based: TinyLlama-1.1B, Qwen3-1.7B; SSM-based: Mamba-1.4B, Mamba-2-1.3B) across three GLUE tasks (SST-2, QNLI, STS-B) and three LaMP personalization tasks (LaMP-1, LaMP-2, LaMP-3). Each configuration is evaluated with the energy-focused NetScore-E and the memory-focused NetScore-M, the two variants that reflect the constraints binding on-device deployment. Methods are selected with a strict energy-first rule (highest NetScore-E, ties broken by NetScore#). LoRA+ achieves the highest NetScore-E in 19 of 24 configurations and the highest NetScore-M in 13 of 24, and is the selected method in 18 of 24. QLoRA, available only for the Transformer models, cuts peak finetuning VRAM by up to 3.9x relative to LoRA and therefore takes the best NetScore-M in 5 of the 12 Transformer configurations, although its de-quantization overhead leaves it selected in only one of them once energy decides. BitFit and full fine-tuning are almost never competitive on either variant, and TinyLlama-1.1B leads the energy-focused NetScore-E on five of the six benchmarks and the memory-focused NetScore-M on four. These results show that compact SLMs paired with PEFT provide a practical, energy-aware path to personalized on-device deployment, with the optimal method set by the dominant constraint: LoRA+ for energy and QLoRA for memory.