Results 221 to 230 of about 9,649,421 (248)

Computational Limits of Low-Rank Adaptation (LoRA) for Transformer-Based Models. [PDF]

CoRR
We study the computational limits of Low-Rank Adaptation (LoRA) update for finetuning transformer-based models using fine-grained complexity theory. Our key observation is that the existence of low-rank decompositions within the gradient computation of ...
Jerry Yao-Chieh Hu, Maojiang Su, En-Jui Kuo, Zhao Song 0002, Han Liu 0001   +4 more
core   +3 more sources

LoRA-NIR: Low-Rank Adaptation of Vision Transformers for Remote Sensing With Near-Infrared Imagery

IEEE Geoscience and Remote Sensing Letters
Plant health can be monitored dynamically using multispectral sensors that measure near-infrared (NIR). Despite this potential, obtaining and annotating high-resolution NIR images pose significant challenges for training deep neural networks.
Erdem Akagündüz, O Özgur Tanriover, Irem Ulku   +2 more
exaly   +2 more sources

Some of the next articles are maybe not open access.

Related searches:

L2-LoRA: Improving Low-Rank Adaptation with Layer-Specific Regularization

Proceedings of the AAAI Conference on Artificial Intelligence
Fine-tuning large language models (LLMs) in a parameter-efficient manner while preserving their pre-trained world knowledge remains a significant challenge. While Low-Rank Adaptation (LoRA) and its variants effectively mitigate catastrophic forgetting, they do not fully eliminate the loss of critical pre-trained knowledge.
Xiang Zhang, Rui Xie 0003, Shikun Zhang
openaire   +1 more source

LoRA-SL: Low-Rank Adaptation for Continual Split Learning

Anais do XLIV Simpósio Brasileiro de Redes de Computadores e Sistemas Distribuídos (SBRC 2026)
In scenarios with drones on multiple missions and diverse tasks, it is common to use multiple servers training data models for different tasks. However, directly training a model on different tasks can lead to catastrophic forgetting, impairing model accuracy on old tasks and requiring retraining.
Mateus C. Oliveira, Heitor H. da Silva, Camilo H. M. dos Santos, Carlos Senna, Allan M. de Souza, Luiz F. Bittencourt   +5 more
openaire   +1 more source

S²LoRA: Subspace-Constrained Low-Rank Adaptation for Deepfake Detection

With the rapid advancement of deepfake generation techniques, deepfake detection has become a critical task in computer vision. However, existing models often exhibit limited generalization capability across datasets and in real-world scenarios, particularly when handling diverse forgery methods.
Zhenrong Deng, Rui Yang, Jinjie Yan, Sheng Liu   +3 more
openaire   +1 more source

LoRA-E ² : Effective and Efficient Low-rank Adaptation

Proceedings of the ACM Web Conference 2026
Shengkun Zhu, Jinshan Zeng, Yiming Wang, Sheng Wang, Yuan Sun, Shangfeng Chen, Yuan Yao, Qiang Yang   +7 more
openaire   +2 more sources

Leech-LoRA: Low-Rank Lattice Adaptation for Large Language Models

We introduce Leech-LoRA, a parameter-efficient fine-tuning method that injects geometric priors from the Leech lattice into large pre-trained Transformer models. Unlike standard LoRA which adds trainable low-rank matrices, Leech-LoRA adds a parallel path through a fixed orthogonal matrix derived from the Leech lattice’s 24-dimensional basis, scaled by
openaire   +1 more source

FI-LoRA: Dynamic and Lightweight Low-Rank Adaptation Guided by Fisher Information

IEEE Signal Processing Letters
Chaobin Zeng, Shusheng Zhao
openaire   +1 more source

G-LoRA: Global-Local Decoupled Low-Rank Adaptation

Findings of the Association for Computational Linguistics: ACL 2026
Jiahao Xiong, Yihong Huang, Yihe Liu, Xianming Hu, Hongbo Zhao, Kai Zhang   +5 more
openaire   +1 more source

ARD-LoRA: Dynamic Rank Allocation for Parameter-Efficient Fine-Tuning of Foundation Models with Heterogeneous Adaptation Needs

IEEE Transactions on Artificial Intelligence
Muhammad Usama
exaly