arXiv Computer Vision September 28, 2026 By Sehyun Kwon

Preserve-and-Compose Training for Composed Image Retrieval

Read the original on arXiv Computer Vision →

The paper introduces Preserve-and-Compose Training (PACT) for composed image retrieval, a task where a query image is modified by a textual instruction while preserving visual content from a reference image. PACT learns from image–text–text triplets, using target captions for supervision and visual evidence from the source image to maintain relevant details, without requiring target images or gallery updates. The authors also propose Chord scoring, which blends target similarity with source-relative directional agreement in a frozen image space, and demonstrate that this combined approach yields strong retrieval performance across multiple zero-shot CIR benchmarks and various backbones.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Computer Vision.

benchmarks

Hugging Face Trending Papers

Jul 6

DiCE-CIR: Direct Composition Learning for Efficient Zero-Shot Composed Image Retrieval

Zero-shot composed image retrieval (ZS-CIR) aims to retrieve a target image from a multimodal query consisting of a reference image and an edit text describing the desired modification. Recent ZS-CIR studies have relied on projection-based methods that map a reference image into pseudo-word tokens in the text embedding space.

llms rag multimodal benchmarks safety

arXiv AI

Jul 2

Learning to Compose: Revisiting Proxy Task Design for Zero-Shot Composed Image Retrieval

arXiv:2607. 00374v1 Announce Type: cross Abstract: Composed Image Retrieval (CIR) retrieves a target image from a reference image and a textual modification.

By Jingjing Zhang, Lei Zhang, Zheren Fu, Zhendong Mao

multimodal benchmarks

arXiv AI

Jun 8

Never Seen Before: Benchmarking Genuine Zero-Shot Composed Image Retrieval with Consistent Video-Sourced Datasets

arXiv:2606. 07032v1 Announce Type: cross Abstract: Zero-Shot Composed Image Retrieval (ZS-CIR) aims to retrieve a target image based on a query composed of a reference image and a relative caption without training samples.

By Zhenyu Yang, Zemin Du, Shengsheng Qian, Changsheng Xu

llms benchmarks

arXiv Computer Vision

Aug 25

Training-Free Pseudo-Fusion for Composed Image Retrieval with Diffusion Models and Multimodal Large Language Models

arXiv:2608.23102v1 Announce Type: new Abstract: Composed Image Retrieval (CIR) is an emerging paradigm in content-based image retrieval that enables users to formulate compositional queries by combin...

By Fan Xu, Luis A. Leiva

llms rag diffusion multimodal benchmarks

arXiv AI

Jun 16

Cross-modal Identity Mapping: Minimizing Information Loss in Modality Conversion via Reinforcement Learning

arXiv:2603. 01696v2 Announce Type: replace-cross Abstract: Large Vision-Language Models (LVLMs) often omit or misrepresent critical visual content in generated image captions.

By Haonan Jia, Shichao Dong, Xin Dong, Zenghui Sun, Jin Wang, Jinsong Lan, Xiaoyong Zhu, Bo Zheng, Kaifu Zhang

llms reinforcement-learning fine-tuning multimodal benchmarks

arXiv AI

Jun 8

TEVI: Text-Conditioned Editing of Visual Representations via Sparse Autoencoders for Improved Vision-Language Alignment

arXiv:2606. 07451v1 Announce Type: cross Abstract: Vision-language models such as CLIP are highly useful for diverse tasks due to their shared image-text embedding space.

By Sweta Mahajan, Sukrut Rao, Jiahao Xie, Alexander Koller, Bernt Schiele

llms rag multimodal benchmarks safety

One email a morning, machine-written

Related stories

DiCE-CIR: Direct Composition Learning for Efficient Zero-Shot Composed Image Retrieval

Learning to Compose: Revisiting Proxy Task Design for Zero-Shot Composed Image Retrieval

Never Seen Before: Benchmarking Genuine Zero-Shot Composed Image Retrieval with Consistent Video-Sourced Datasets

Training-Free Pseudo-Fusion for Composed Image Retrieval with Diffusion Models and Multimodal Large Language Models

Cross-modal Identity Mapping: Minimizing Information Loss in Modality Conversion via Reinforcement Learning

TEVI: Text-Conditioned Editing of Visual Representations via Sparse Autoencoders for Improved Vision-Language Alignment