arXiv 2602.044512026-02-05cs.IR

SDR-CIR：一种基于链式推理的语义去偏检索框架用于无训练零样本复合图像检索

SDR-CIR: Semantic Debias Retrieval Framework for Training-Free Zero-Shot Composed Image Retrieval

Wuhan University of Technology（武汉理工大学）；Wuhan Vocational College of Software and Engineering（武汉软件工程职业学院）

Yi Sun, Jinyu Xu, Qing Xie, Jiachen Li, Yanchun Ma, Yongjian Liu

备注：: Accepted by WWW 2026

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

SDR-CIR通过语义去偏排名方法提升无训练零样本复合图像检索性能。

摘要：

复合图像检索（CIR）旨在从由参考图像和修改文本组成的查询中检索目标图像。最近的无训练零样本方法通常使用多模态大语言模型（MLLM）结合链式推理（CoT）来生成目标图像描述以供检索。然而，由于ZS-CIR的模糊匹配性质，生成的描述容易相对于目标图像产生语义偏差。我们提出了SDR-CIR，一种基于链式推理的语义去偏排名方法。首先，选择性CoT引导MLLM在图像理解过程中提取与修改文本相关的视觉内容，从而在源头上减少视觉噪声。然后，我们引入了两步的语义去偏排名，即锚定和去偏。在锚定步骤中，我们将参考图像特征与目标描述特征融合，以强化有用的语义并补充缺失的提示。在去偏步骤中，我们显式建模参考图像对描述的视觉语义贡献，并将其纳入相似度分数作为惩罚项。通过补充缺失的提示同时抑制冗余，SDR-CIR缓解了语义偏差并提高了检索性能。在三个标准CIR基准上的实验表明，SDR-CIR在单阶段方法中取得了最先进的结果，同时保持了高效率。代码已公开在https://github.com/suny105/SDR-CIR。

英文摘要：

Composed Image Retrieval (CIR) aims to retrieve a target image from a query composed of a reference image and modification text. Recent training-free zero-shot methods often employ Multimodal Large Language Models (MLLMs) with Chain-of-Thought (CoT) to compose a target image description for retrieval. However, due to the fuzzy matching nature of ZS-CIR, the generated description is prone to semantic bias relative to the target image. We propose SDR-CIR, a training-free Semantic Debias Ranking method based on CoT reasoning. First, Selective CoT guides the MLLM to extract visual content relevant to the modification text during image understanding, thereby reducing visual noise at the source. We then introduce a Semantic Debias Ranking with two steps, Anchor and Debias, to mitigate semantic bias. In the Anchor step, we fuse reference image features with target description features to reinforce useful semantics and supplement omitted cues. In the Debias step, we explicitly model the visual semantic contribution of the reference image to the description and incorporate it into the similarity score as a penalty term. By supplementing omitted cues while suppressing redundancy, SDR-CIR mitigates semantic bias and improves retrieval performance. Experiments on three standard CIR benchmarks show that SDR-CIR achieves state-of-the-art results among one-stage methods while maintaining high efficiency. The code is publicly available at https://github.com/suny105/SDR-CIR.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。