Dual-Path Stable Soft Prompt Generation for Domain Generalization

TL;DR

This paper introduces DPSPG, a transformer-based framework that uses negative learning to generate stable, effective prompts for domain generalization, outperforming existing methods across multiple benchmarks.

Contribution

It proposes a novel dual-path soft prompt generation method incorporating negative learning to enhance prompt stability and generalization in domain generalization tasks.

Findings

DPSPG outperforms state-of-the-art methods on five benchmark datasets.

Negative learning improves prompt robustness and reduces variability.

Theoretical analysis confirms increased margin and reduced gradient norm upper bound.

Abstract

Domain generalization (DG) aims to learn a model using data from one or multiple related but distinct source domains that can generalize well to unseen out-of-distribution target domains. Inspired by the success of large pre-trained vision-language models (VLMs), prompt tuning has emerged as an effective generalization strategy. However, it often struggles to capture domain-specific features due to its reliance on manually or fixed prompt inputs. Recently, some prompt generation methods have addressed this limitation by dynamically generating instance-specific and domain-specific prompts for each input, enriching domain information and demonstrating potential for enhanced generalization. Through further investigation, we identify a notable issue in existing prompt generation methods: the same input often yields significantly different and suboptimal prompts across different random seeds, a phenomenon we term Prompt Variability. To address this, we introduce negative learning into the prompt generation process and propose Dual-Path Stable Soft Prompt Generation (DPSPG), a transformer-based framework designed to improve both the stability and generalization of prompts. Specifically, DPSPG incorporates a complementary prompt generator to produce negative prompts, thereby reducing the risk of introducing misleading information. Both theoretical and empirical analyses demonstrate that negative learning leads to more robust and effective prompts by increasing the effective margin and reducing the upper bound of the gradient norm. Extensive experiments on five DG benchmark datasets show that DPSPG consistently outperforms state-of-the-art methods while maintaining prompt stability.