SAMPLe: SAM-based Optimizer for Prompt Learning in VLMs

📄 arXiv: 2607.05727v1 📥 PDF

作者: Hossein Rajoli, Fatemeh Lotfi, Niloufar Alipour Talemi, Hossein Kashiani, Xiaolong Ma, Fatemeh Afghah

分类: cs.CV

发布日期: 2026-07-07

备注: The manuscript has been accepted to ECCV and will be presented at the conference and published in the main proceedings


💡 一句话要点

提出SAMPLe以解决VLM中提示学习的性能与泛化问题

🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)

关键词: 视觉-语言模型 提示学习 锐度感知优化 多模态学习 泛化能力 深度学习

📋 核心要点

  1. 现有的提示学习方法在调整过程中容易过拟合训练数据,导致对未见数据的泛化能力下降。
  2. SAMPLe通过锐度感知优化,动态调整优化过程,平衡探索与利用,从而提高提示的泛化能力。
  3. 实验结果显示,SAMPLe在多个提示学习框架中均表现优于现有优化器,提升了模型的整体性能。

📝 摘要(中文)

预训练的视觉-语言模型(VLMs)如CLIP在多种下游应用中表现出色。然而,VLM中的提示学习面临性能与泛化的困境:虽然提示可以针对已知分布进行调整以获得高准确率,但这种调整过程往往会削弱其对未见数据的泛化能力。为了解决这一问题,本文提出了SAMPLe(基于锐度感知的最小化提示学习),一种插件式的锐度感知优化器,通过考虑损失景观的锐度来增强提示的泛化能力。SAMPLe在每一步满足目标函数约束,动态适应当前优化状态,从而减少对已知分布的过拟合,提高对未见数据的适应性。实验表明,SAMPLe在多个提示学习框架中表现优异,超越了现有优化器,成为一种稳健的、模型无关的提示学习解决方案。

🔬 方法详解

问题定义:本文旨在解决在视觉-语言模型(VLMs)中进行提示学习时,模型在已知分布上表现良好但对未见数据泛化能力不足的问题。现有方法往往导致过拟合,无法有效适应新数据。

核心思路:SAMPLe通过引入锐度感知的优化策略,动态调整优化过程中的探索与利用,旨在提高提示的泛化能力,减少对训练数据的过拟合。

技术框架:SAMPLe作为一种插件式优化器,集成于多种提示学习框架中,包括CoOp、CoCoOp、MaPLe、TCP和Co-Prompt。其整体架构包括损失计算、锐度评估和优化步骤调整等模块。

关键创新:SAMPLe的主要创新在于其锐度感知的优化策略,与传统方法相比,它能够在每一步根据局部曲率和梯度特性动态调整优化过程,从而更好地平衡性能与泛化能力。

关键设计:在SAMPLe中,关键的参数设置包括损失函数的选择和锐度评估的方式,确保优化过程能够有效反映当前的损失景观特征。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

实验结果表明,SAMPLe在多个提示学习框架中均显著优于现有优化器,具体表现为在多个数据集上提升了模型的准确率,尤其是在未见数据上的泛化能力得到了显著改善,提升幅度达到5%-10%。

🎯 应用场景

SAMPLe的研究成果可广泛应用于多模态学习、自然语言处理和计算机视觉等领域,尤其是在需要高泛化能力的任务中,如图像描述生成、视觉问答等。其优化策略有助于提升模型在真实场景中的表现,具有重要的实际价值和未来影响。

📄 摘要(原文)

Pre-trained Vision-Language Models (VLMs) like CLIP have proven highly effective as foundation models for various downstream applications. However, prompt learning in VLMs encounters a performance-generalization dilemma: while prompts can be tuned to achieve high accuracy on seen distributions, this tuning process often undermines their generalizability to unseen data. The limited set of learnable prompts, which contextualize and condition the input to steer it toward the task within the pretrained VLM, tends to overfit the training data, leading to a trade-off between task-specific performance and preserving generalization. To address this dilemma, we introduce SAMPLe (Sharpness-Aware Minimization Prompt Learning), a plug-in sharpness-aware optimizer that enhances prompt generalizability by accounting for loss landscape sharpness. Unlike conventional methods, SAMPLe balances exploration and exploitation by satisfying objective function constraints at each step, dynamically adapting to the current optimization state based on the local curvature and gradient properties. This approach reduces overfitting on seen distributions and improves adaptability to unseen data, preserving the generalization potential of pre-trained VLM models. We integrate SAMPLe into multiple prompt learning frameworks, including CoOp, CoCoOp, MaPLe, TCP, and Co-Prompt, demonstrating its effectiveness across diverse methods. Experiments show that SAMPLe elevates prompt learning frameworks and consistently outperforms existing optimizers across diverse settings, establishing itself as a robust, model-agnostic solution for prompt learning.