Saved in:
Bibliographic Details
Main Authors: Wan, Guoan, Chen, Tianyu, Feng, Fangzheng, Zhou, Haoyi, Xu, Runhua
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2512.23485
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914223850782720
author Wan, Guoan
Chen, Tianyu
Feng, Fangzheng
Zhou, Haoyi
Xu, Runhua
author_facet Wan, Guoan
Chen, Tianyu
Feng, Fangzheng
Zhou, Haoyi
Xu, Runhua
contents Parameter-efficient fine-tuning (PEFT) methods have emerged as a practical solution for adapting large foundation models to downstream tasks, reducing computational and memory costs by updating only a small subset of parameters. Among them, approaches like LoRA aim to strike a balance between efficiency and expressiveness, but often suffer from slow convergence and limited adaptation capacity due to their inherent low-rank constraints. This trade-off hampers the ability of PEFT methods to capture complex patterns needed for diverse tasks. To address these challenges, we propose FRoD, a novel fine-tuning method that combines hierarchical joint decomposition with rotational degrees of freedom. By extracting a globally shared basis across layers and injecting sparse, learnable perturbations into scaling factors for flexible full-rank updates, FRoD enhances expressiveness and efficiency, leading to faster and more robust convergence. On 20 benchmarks spanning vision, reasoning, and language understanding, FRoD matches full model fine-tuning in accuracy, while using only 1.72% of trainable parameters under identical training budgets.
format Preprint
id arxiv_https___arxiv_org_abs_2512_23485
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle FRoD: Full-Rank Efficient Fine-Tuning with Rotational Degrees for Fast Convergence
Wan, Guoan
Chen, Tianyu
Feng, Fangzheng
Zhou, Haoyi
Xu, Runhua
Machine Learning
Artificial Intelligence
Parameter-efficient fine-tuning (PEFT) methods have emerged as a practical solution for adapting large foundation models to downstream tasks, reducing computational and memory costs by updating only a small subset of parameters. Among them, approaches like LoRA aim to strike a balance between efficiency and expressiveness, but often suffer from slow convergence and limited adaptation capacity due to their inherent low-rank constraints. This trade-off hampers the ability of PEFT methods to capture complex patterns needed for diverse tasks. To address these challenges, we propose FRoD, a novel fine-tuning method that combines hierarchical joint decomposition with rotational degrees of freedom. By extracting a globally shared basis across layers and injecting sparse, learnable perturbations into scaling factors for flexible full-rank updates, FRoD enhances expressiveness and efficiency, leading to faster and more robust convergence. On 20 benchmarks spanning vision, reasoning, and language understanding, FRoD matches full model fine-tuning in accuracy, while using only 1.72% of trainable parameters under identical training budgets.
title FRoD: Full-Rank Efficient Fine-Tuning with Rotational Degrees for Fast Convergence
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2512.23485