DeFT-LoRA: Decoupled and Fused Tuning with LoRA Experts for Universal Cross-Domain Retrieval

Ke Xu; Xiaozheng Shen; Shanshan Wang; Mengzhu Wang; Xun Yang

AAAI 2026technical0 citations

DeFT-LoRA: Decoupled and Fused Tuning with LoRA Experts for Universal Cross-Domain Retrieval

Ke Xu, Xiaozheng Shen, Shanshan Wang, Mengzhu Wang, Xun Yang

Abstract

Universal Cross-Domain Retrieval (UCDR) aims to retrieve images across unseen domains and categories, a critical capability for real-world applications. While large-scale Vision-Language Models (VLMs) like CLIP offer strong zero-shot category generalization, they struggle with domain shifts. Existing methods often improve domain robustness at the cost of high computational overhead or by compromising the VLM

BibTeX

@inproceedings{aaai2026_deftloradecouple,
  title = {DeFT-LoRA: Decoupled and Fused Tuning with LoRA Experts for Universal Cross-Domain Retrieval},
  author = {Ke Xu and Xiaozheng Shen and Shanshan Wang and Mengzhu Wang and Xun Yang},
  booktitle = {AAAI 2026},
  year = {2026}
}