AI Navigate

低ランク分解に基づくパラメータ効率的ファインチューニングにおける壊滅的忘却について

arXiv cs.LG / 2026/3/11

Ideas & Deep AnalysisModels & Research

要点

  • LoRAのような低ランク分解を用いたパラメータ効率的ファインチューニング(PEFT)手法は、大規模事前学習モデルの適応に一般的に用いられているが、連続学習における特に壊滅的忘却に関する挙動は十分に理解されていない。
  • 本研究は、忘却がファインチューニング時に用いられる更新部分空間の幾何学およびパラメータ化によって大きく影響されることを実証的に示している。
  • 小さく共有された行列部分空間への更新に制限する手法は、タスク間干渉と忘却を招きやすい。
  • LoRETTAのようなテンソル分解は、超コンパクトなパラメータ予算内で豊かな構造情報を捉えることで忘却の軽減に寄与する。
  • WeGeFTのような構造的に整合したパラメータ化は、連続学習シナリオにおいて事前学習表現をよりよく保持し、更新部分空間の設計が継続学習において重要であることを示している。

Computer Science > Machine Learning

arXiv:2603.09684 (cs)
[Submitted on 10 Mar 2026]

Title:On Catastrophic Forgetting in Low-Rank Decomposition-Based Parameter-Efficient Fine-Tuning

View a PDF of the paper titled On Catastrophic Forgetting in Low-Rank Decomposition-Based Parameter-Efficient Fine-Tuning, by Muhammad Ahmad and Jingjing Zheng and Yankai Cao
View PDF HTML (experimental)
Abstract:Parameter-efficient fine-tuning (PEFT) based on low-rank decomposition, such as LoRA, has become a standard for adapting large pretrained models. However, its behavior in sequential learning -- specifically regarding catastrophic forgetting -- remains insufficiently understood. In this work, we present an empirical study showing that forgetting is strongly influenced by the geometry and parameterization of the update subspace. While methods that restrict updates to small, shared matrix subspaces often suffer from task interference, tensor-based decompositions (e.g., LoRETTA) mitigate forgetting by capturing richer structural information within ultra-compact budgets, and structurally aligned parameterizations (e.g., WeGeFT) preserve pretrained representations. Our findings highlight update subspace design as a key factor in continual learning and offer practical guidance for selecting efficient adaptation strategies in sequential settings.
Subjects: Machine Learning (cs.LG)
Cite as: arXiv:2603.09684 [cs.LG]
  (or arXiv:2603.09684v1 [cs.LG] for this version)
  https://doi.org/10.48550/arXiv.2603.09684
Focus to learn more
arXiv-issued DOI via DataCite

Submission history

From: Jingjing Zheng [view email]
[v1] Tue, 10 Mar 2026 13:53:25 UTC (923 KB)
Full-text links:

Access Paper:

    View a PDF of the paper titled On Catastrophic Forgetting in Low-Rank Decomposition-Based Parameter-Efficient Fine-Tuning, by Muhammad Ahmad and Jingjing Zheng and Yankai Cao
  • View PDF
  • HTML (experimental)
  • TeX Source
Current browse context:
cs.LG
< prev   |   next >
Change to browse by:
cs

References & Citations

export BibTeX citation Loading...

BibTeX formatted citation

×
Data provided by:

Bookmark

BibSonomy logo Reddit logo
Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle
Bibliographic Explorer (What is the Explorer?)
Connected Papers Toggle
Connected Papers (What is Connected Papers?)
Litmaps Toggle
Litmaps (What is Litmaps?)
scite.ai Toggle
scite Smart Citations (What are Smart Citations?)
Code, Data, Media

Code, Data and Media Associated with this Article

alphaXiv Toggle
alphaXiv (What is alphaXiv?)
Links to Code Toggle
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub Toggle
DagsHub (What is DagsHub?)
GotitPub Toggle
Gotit.pub (What is GotitPub?)
Huggingface Toggle
Hugging Face (What is Huggingface?)
Links to Code Toggle
Papers with Code (What is Papers with Code?)
ScienceCast Toggle
ScienceCast (What is ScienceCast?)
Demos

Demos

Replicate Toggle
Replicate (What is Replicate?)
Spaces Toggle
Hugging Face Spaces (What is Spaces?)
Spaces Toggle
TXYZ.AI (What is TXYZ.AI?)
Related Papers

Recommenders and Search Tools

Link to Influence Flower
Influence Flower (What are Influence Flowers?)
Core recommender toggle
CORE Recommender (What is CORE?)
IArxiv recommender toggle
IArxiv Recommender (What is IArxiv?)
About arXivLabs

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.