LoFT: Low-Rank Adaptation That Behaves Like Full Fine-Tuning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Tastan, Nurbek, Laskaridis, Stefanos, Takac, Martin, Nandakumar, Karthik, Horvath, Samuel
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910045486186496
author Tastan, Nurbek
Laskaridis, Stefanos
Takac, Martin
Nandakumar, Karthik
Horvath, Samuel
author_facet Tastan, Nurbek
Laskaridis, Stefanos
Takac, Martin
Nandakumar, Karthik
Horvath, Samuel
contents Large pre-trained models are commonly adapted to downstream tasks using parameter-efficient fine-tuning methods such as Low-Rank Adaptation (LoRA), which injects small trainable low-rank matrices instead of updating all weights. While LoRA dramatically reduces trainable parameters with little overhead, it can still underperform full fine-tuning in accuracy and often converges more slowly. We introduce LoFT, a novel low-rank adaptation method that behaves like full fine-tuning by aligning the optimizer's internal dynamics with those of updating all model weights. LoFT not only learns weight updates in a low-rank subspace (like LoRA) but also properly projects the optimizer's first and second moments (Adam's momentum and variance) into the same subspace, mirroring full-model updates. By aligning the low-rank update itself with the full update, LoFT eliminates the need for tuning extra hyperparameters, e.g., the LoRA scaling factor $α$. Empirically, this approach substantially narrows the performance gap between adapter-based tuning and full fine-tuning and consistently outperforms standard LoRA-style methods, all without increasing inference cost.
format Preprint
id arxiv_https___arxiv_org_abs_2505_21289
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle LoFT: Low-Rank Adaptation That Behaves Like Full Fine-Tuning
Tastan, Nurbek
Laskaridis, Stefanos
Takac, Martin
Nandakumar, Karthik
Horvath, Samuel
Machine Learning
Optimization and Control
Large pre-trained models are commonly adapted to downstream tasks using parameter-efficient fine-tuning methods such as Low-Rank Adaptation (LoRA), which injects small trainable low-rank matrices instead of updating all weights. While LoRA dramatically reduces trainable parameters with little overhead, it can still underperform full fine-tuning in accuracy and often converges more slowly. We introduce LoFT, a novel low-rank adaptation method that behaves like full fine-tuning by aligning the optimizer's internal dynamics with those of updating all model weights. LoFT not only learns weight updates in a low-rank subspace (like LoRA) but also properly projects the optimizer's first and second moments (Adam's momentum and variance) into the same subspace, mirroring full-model updates. By aligning the low-rank update itself with the full update, LoFT eliminates the need for tuning extra hyperparameters, e.g., the LoRA scaling factor $α$. Empirically, this approach substantially narrows the performance gap between adapter-based tuning and full fine-tuning and consistently outperforms standard LoRA-style methods, all without increasing inference cost.
title LoFT: Low-Rank Adaptation That Behaves Like Full Fine-Tuning
topic Machine Learning
Optimization and Control
url https://arxiv.org/abs/2505.21289