# Preserve-and-Compose Training for Composed Image Retrieval

arXiv: https://arxiv.org/abs/2609.31202 · HTML: https://mlsanity.com/paper/2609.31202/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 21 of its 33 references that are in the map.
- First posted on arXiv 25 Sep 2026.

## Builds on (most cited first, 21 of 21)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [LAION-5B: An open large-scale dataset for training next generation image-text models](https://mlsanity.com/paper/2210.08402/) (2022)
- [Reproducible scaling laws for contrastive language-image learning](https://mlsanity.com/paper/2212.07143/) (2022)
- [Image Retrieval on Real-life Images with Pre-trained Vision-and-Language Models](https://mlsanity.com/paper/2108.04024/) (2021)
- [MovieNet: A Holistic Dataset for Movie Understanding](https://mlsanity.com/paper/2007.10937/) (2020)
- [Zero-Shot Composed Image Retrieval with Textual Inversion](https://mlsanity.com/paper/2303.15247/) (2023)
- [Pic2Word: Mapping Pictures to Words for Zero-shot Composed Image Retrieval](https://mlsanity.com/paper/2302.03084/) (2023)
- [Scaling Deep Contrastive Learning Batch Size under Memory Limited Setup](https://mlsanity.com/paper/2101.06983/) (2021)
- [Vision-by-Language for Training-Free Compositional Image Retrieval](https://mlsanity.com/paper/2310.09291/) (2023)
- [DreamLIP: Language-Image Pre-training with Long Captions](https://mlsanity.com/paper/2403.17007/) (2024)
- [Context-I2W: Mapping Images to Context-dependent Words for Accurate Zero-Shot Composed Image Retrieval](https://mlsanity.com/paper/2309.16137/) (2023)
- [GeneCIS: A Benchmark for General Conditional Image Similarity](https://mlsanity.com/paper/2306.07969/) (2023)
- [Spherical Linear Interpolation and Text-Anchoring for Zero-shot Composed Image Retrieval](https://mlsanity.com/paper/2405.00571/) (2024)
- [CoLLM: A Large Language Model for Composed Image Retrieval](https://mlsanity.com/paper/2503.19910/) (2025)
- [Missing Target-Relevant Information Prediction with World Model for Accurate Zero-Shot Composed Image Retrieval](https://mlsanity.com/paper/2503.17109/) (2025)
- [PDV: Prompt Directional Vectors for Zero-shot Composed Image Retrieval](https://mlsanity.com/paper/2502.07215/) (2025)
- [An Efficient Post-hoc Framework for Reducing Task Discrepancy of Text Encoders for Composed Image Retrieval](https://mlsanity.com/paper/2406.09188/) (2024)
- [Modality and Task Adaptation for Enhanced Zero-shot Composed Image Retrieval](https://mlsanity.com/paper/2410.23736/) (2024)
- [FlowCIR: Semantic Transport via Flow Matching for Zero-Shot Composed Image Retrieval](https://mlsanity.com/paper/2607.02284/) (2026)
- [Learning to Compose: Revisiting Proxy Task Design for Zero-Shot Composed Image Retrieval](https://mlsanity.com/paper/2607.00374/) (2026)

## Abstract (arXiv)

Composed image retrieval (CIR) aims to retrieve images that satisfy a user-specified modification while preserving relevant visual content from a reference image. Collecting target images for this purpose is costly, motivating zero-shot CIR methods that instead use target captions as supervision. However, target captions may omit source details that should be preserved. We therefore propose, Preserve-and-Compose Training, which complements target-caption supervision with visual evidence from the source image. PACT learns from image--text--text (ITT) triplets without target images or gallery updates, aligning composed queries with target captions while preserving source evidence through visual supervision. We further introduce Chord scoring, which combines target similarity with source-relative directional agreement in the frozen image space. Results across four ZS-CIR benchmarks show that combining target-caption supervision with source-image evidence leads to strong retrieval performance across datasets, backbone scales, and external galleries. The code is available on https://github.com/sehyunkwon/PACT.