Decoupling Augmentation Bias in Prompt Learning for Vision-Language Models

TL;DR

This paper introduces AAPL, a novel method that uses adversarial token embeddings to decouple augmentation bias from semantic features in prompt learning, significantly improving generalization in vision-language models.

Contribution

AAPL is the first approach to explicitly separate augmentation-induced superficial variations from semantic features in prompt learning for vision-language models.

Findings

AAPL outperforms existing methods across multiple benchmarks.

Image-based augmentation enhances prompt learning effectiveness.

Decoupling improves generalization in few-shot and zero-shot tasks.

Abstract

Recent advances in large-scale vision and language models have led to significant progress in zero-shot learning tasks. Methods such as CoOp and CoCoOp have shown that replacing handcrafted prompts with learnable vectors, known as prompt learning, can result in improved performance. However, these models often struggle to generalize to entirely unseen categories. While traditional zero-shot learning techniques benefit from various data augmentation strategies, prompt learning has primarily focused on text-based modifications, leaving the potential of image-based augmentation largely unexplored. In this work, we explore how image-level augmentations, particularly those that introduce attribute-specific variations, can support and enhance prompt learning. Our analysis examines the interaction between these augmentations and soft prompt frameworks, revealing their potential to improve generalization. We also identify a limitation in existing methods, such as CoCoOp, which do not provide explicit guidance for learning prompts that focus on semantically meaningful visual features. To address this, we propose Adding Attributes to Prompt Learning, AAPL, a novel method that introduces adversarial token embeddings to decouple superficial visual variations introduced by augmentation from class-relevant semantic representations. This decoupling enables the learned prompts to concentrate on visually discriminative features that align with the target categories. We conduct comprehensive experiments on eleven benchmark datasets, and AAPL consistently outperforms existing methods across few-shot, zero-shot, cross-dataset, and domain generalization settings. Our source code is publicly available at: https://github.com/Gahyeonkim09/AAPL