LoRA-Gen: Specializing Large Language Model via Online LoRA Generation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Xiao, Yicheng, Song, Lin, Yang, Rui, Cheng, Cheng, Ge, Yixiao, Li, Xiu, Shan, Ying
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912428484198400
author Xiao, Yicheng
Song, Lin
Yang, Rui
Cheng, Cheng
Ge, Yixiao
Li, Xiu
Shan, Ying
author_facet Xiao, Yicheng
Song, Lin
Yang, Rui
Cheng, Cheng
Ge, Yixiao
Li, Xiu
Shan, Ying
contents Recent advances have highlighted the benefits of scaling language models to enhance performance across a wide range of NLP tasks. However, these approaches still face limitations in effectiveness and efficiency when applied to domain-specific tasks, particularly for small edge-side models. We propose the LoRA-Gen framework, which utilizes a large cloud-side model to generate LoRA parameters for edge-side models based on task descriptions. By employing the reparameterization technique, we merge the LoRA parameters into the edge-side model to achieve flexible specialization. Our method facilitates knowledge transfer between models while significantly improving the inference efficiency of the specialized model by reducing the input context length. Without specialized training, LoRA-Gen outperforms conventional LoRA fine-tuning, which achieves competitive accuracy and a 2.1x speedup with TinyLLaMA-1.1B in reasoning tasks. Besides, our method delivers a compression ratio of 10.1x with Gemma-2B on intelligent agent tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2506_11638
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle LoRA-Gen: Specializing Large Language Model via Online LoRA Generation
Xiao, Yicheng
Song, Lin
Yang, Rui
Cheng, Cheng
Ge, Yixiao
Li, Xiu
Shan, Ying
Computation and Language
Artificial Intelligence
Recent advances have highlighted the benefits of scaling language models to enhance performance across a wide range of NLP tasks. However, these approaches still face limitations in effectiveness and efficiency when applied to domain-specific tasks, particularly for small edge-side models. We propose the LoRA-Gen framework, which utilizes a large cloud-side model to generate LoRA parameters for edge-side models based on task descriptions. By employing the reparameterization technique, we merge the LoRA parameters into the edge-side model to achieve flexible specialization. Our method facilitates knowledge transfer between models while significantly improving the inference efficiency of the specialized model by reducing the input context length. Without specialized training, LoRA-Gen outperforms conventional LoRA fine-tuning, which achieves competitive accuracy and a 2.1x speedup with TinyLLaMA-1.1B in reasoning tasks. Besides, our method delivers a compression ratio of 10.1x with Gemma-2B on intelligent agent tasks.
title LoRA-Gen: Specializing Large Language Model via Online LoRA Generation
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2506.11638