Scholarlink

<p>Pre-trained vision-language models&#xa0;(<i>e.g</i>., CLIP) have shown impressive success in various computer vision tasks with their generalization capability. Recently, parameter-efficient fine-tuning&#xa0;(PEFT) approaches have been actively explored to effectively and efficiently adapt the pre-trained vision-language models to a variety of downstream tasks. However, most existing PEFT approaches suffer from a task overfitting issue since the general knowledge of the pre-trained models is forgotten while a small number of learnable parameters in soft prompts/adapters are fine-tuned on a small data set from a specific target task. Thus, we propose a <b>P</b>arameter-<b>E</b>fficient <b>F</b>ine-<b>T</b>uning via <b>Meta</b>-<b>R</b>egularization&#xa0;(PEFT-MetaR) to improve the generalizability of parameter-efficient fine-tuning methods for vision-language models. Specifically, PEFT-MetaR meta-learns both the regularizer and learnable parameters to harness the task-specific knowledge from the downstream tasks and task-agnostic general knowledge from the pretrained models. Further, PEFT-MetaR augments the task to generate multiple virtual tasks to alleviate the meta-overfitting. In addition, we provide the analysis to comprehend how PEFT-MetaR improves the generalizability from the perspective of the gradient alignment. Our experiments demonstrate that PEFT-MetaR improves the generalizability of parameter-efficient fine-tuning methods on various datasets.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Parameter-Efficient Fine-Tuning via Meta-Regularizer

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月6日

查看全文
  • Jinyoung Park,
  • Juyeon Ko,
  • Sanghyeok Lee,
  • Joonmyung Choi,
  • Hyunwoo J. Kim

摘要

Pre-trained vision-language models (e.g., CLIP) have shown impressive success in various computer vision tasks with their generalization capability. Recently, parameter-efficient fine-tuning (PEFT) approaches have been actively explored to effectively and efficiently adapt the pre-trained vision-language models to a variety of downstream tasks. However, most existing PEFT approaches suffer from a task overfitting issue since the general knowledge of the pre-trained models is forgotten while a small number of learnable parameters in soft prompts/adapters are fine-tuned on a small data set from a specific target task. Thus, we propose a Parameter-Efficient Fine-Tuning via Meta-Regularization (PEFT-MetaR) to improve the generalizability of parameter-efficient fine-tuning methods for vision-language models. Specifically, PEFT-MetaR meta-learns both the regularizer and learnable parameters to harness the task-specific knowledge from the downstream tasks and task-agnostic general knowledge from the pretrained models. Further, PEFT-MetaR augments the task to generate multiple virtual tasks to alleviate the meta-overfitting. In addition, we provide the analysis to comprehend how PEFT-MetaR improves the generalizability from the perspective of the gradient alignment. Our experiments demonstrate that PEFT-MetaR improves the generalizability of parameter-efficient fine-tuning methods on various datasets.