Diffusion-Driven Progressive Target Manipulation for Source-Free Domain Adaptation

2025-10-29

AI Summary

To address the unreliability of pseudo-labels and the exacerbation of domain shift caused by pseudo-data generation in Source-Free Domain Adaptation (SFDA), this paper proposes a diffusion-based progressive optimization framework for the target domain. Relying solely on a pre-trained source model and unlabeled target data, our method performs trustworthy sample partitioning, semantic-guided latent diffusion modeling, pseudo-label reliability assessment, and iterative reclassification of untrustworthy samples—enabling high-fidelity pseudo-target data generation and adaptive optimization while preserving the intrinsic target distribution. The core innovation lies in integrating diffusion models into SFDA, establishing a semantics-controllable, progressive mechanism for pseudo-sample generation and refinement. Extensive experiments on four mainstream SFDA benchmarks demonstrate substantial improvements over state-of-the-art methods, with gains of up to 18.6%, achieving new state-of-the-art performance.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• User Modeling, Personalization and Recommendation: User modeling for targeted and personalized online advertising
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Source-free domain adaptation (SFDA) is a challenging task that tackles domain shifts using only a pre-trained source model and unlabeled target data. Existing SFDA methods are restricted by the fundamental limitation of source-target domain discrepancy. Non-generation SFDA methods suffer from unreliable pseudo-labels in challenging scenarios with large domain discrepancies, while generation-based SFDA methods are evidently degraded due to enlarged domain discrepancies in creating pseudo-source data. To address this limitation, we propose a novel generation-based framework named Diffusion-Driven Progressive Target Manipulation (DPTM) that leverages unlabeled target data as references to reliably generate and progressively refine a pseudo-target domain for SFDA. Specifically, we divide the target samples into a trust set and a non-trust set based on the reliability of pseudo-labels to sufficiently and reliably exploit their information. For samples from the non-trust set, we develop a manipulation strategy to semantically transform them into the newly assigned categories, while simultaneously maintaining them in the target distribution via a latent diffusion model. Furthermore, we design a progressive refinement mechanism that progressively reduces the domain discrepancy between the pseudo-target domain and the real target domain via iterative refinement. Experimental results demonstrate that DPTM outperforms existing methods by a large margin and achieves state-of-the-art performance on four prevailing SFDA benchmark datasets with different scales. Remarkably, DPTM can significantly enhance the performance by up to 18.6% in scenarios with large source-target gaps.

Problem

• Addressing unreliable pseudo-labels in source-free domain adaptation
• Reducing domain discrepancies through progressive target manipulation
• Generating refined pseudo-target data using latent diffusion models

Innovation

• Leverages target data to generate pseudo-target domain
• Uses latent diffusion model for semantic transformation
• Progressively refines domain discrepancy via iterative mechanism