Zero-Order Sharpness-Aware Minimization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Fu, Yao, Jin, Yihang, Zhang, Chunxia, Liu, Junmin, Dai, Guang, Ye, Haishan
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915695998009344
author Fu, Yao
Jin, Yihang
Zhang, Chunxia
Liu, Junmin
Dai, Guang
Ye, Haishan
author_facet Fu, Yao
Jin, Yihang
Zhang, Chunxia
Liu, Junmin
Dai, Guang
Ye, Haishan
contents Prompt learning has become a key method for adapting large language models to specific tasks with limited data. However, traditional gradient-based optimization methods for tuning prompts are computationally intensive, posing challenges for efficiency. We introduce ZOSA (Zero-Order Sharpness-Aware Minimization), a novel optimization framework that integrates zero-order optimization with sharpness-aware minimization to enhance prompt tuning. ZOSA employs Rademacher perturbation vectors to estimate gradients without requiring backpropagation. By incorporating sharpness-aware principles, it targets flat minima in the loss landscape, improving generalization. An adaptive learning rate, guided by loss variability, further ensures stable convergence. Experiments on few-shot learning tasks, such as text classification and natural language inference, show that ZOSA significantly outperforms existing methods. With its theoretical foundation and computational efficiency, ZOSA offers a practical solution for prompt-based learning in resource-limited settings.
format Preprint
id arxiv_https___arxiv_org_abs_2511_09156
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Zero-Order Sharpness-Aware Minimization
Fu, Yao
Jin, Yihang
Zhang, Chunxia
Liu, Junmin
Dai, Guang
Ye, Haishan
Statistics Theory
Prompt learning has become a key method for adapting large language models to specific tasks with limited data. However, traditional gradient-based optimization methods for tuning prompts are computationally intensive, posing challenges for efficiency. We introduce ZOSA (Zero-Order Sharpness-Aware Minimization), a novel optimization framework that integrates zero-order optimization with sharpness-aware minimization to enhance prompt tuning. ZOSA employs Rademacher perturbation vectors to estimate gradients without requiring backpropagation. By incorporating sharpness-aware principles, it targets flat minima in the loss landscape, improving generalization. An adaptive learning rate, guided by loss variability, further ensures stable convergence. Experiments on few-shot learning tasks, such as text classification and natural language inference, show that ZOSA significantly outperforms existing methods. With its theoretical foundation and computational efficiency, ZOSA offers a practical solution for prompt-based learning in resource-limited settings.
title Zero-Order Sharpness-Aware Minimization
topic Statistics Theory
url https://arxiv.org/abs/2511.09156