# Weaving Visual Narratives: Agentic Image Bundle Composition Beyond Atomic Visual Matching

arXiv: https://arxiv.org/abs/2608.28695 · HTML: https://mlsanity.com/paper/2608.28695/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 15 of its 34 references that are in the map.
- First posted on arXiv 27 Aug 2026.

## Builds on (most cited first, 15 of 15)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [Microsoft COCO: Common Objects in Context](https://mlsanity.com/paper/1405.0312/) (2014)
- [Flickr30k Entities: Collecting Region-to-Phrase Correspondences for Richer Image-to-Sentence Models](https://mlsanity.com/paper/1505.04870/) (2015)
- [SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features](https://mlsanity.com/paper/2502.14786/) (2025)
- [Qwen3-VL-Embedding and Qwen3-VL-Reranker: A Unified Framework for State-of-the-Art Multimodal Retrieval and Ranking](https://mlsanity.com/paper/2601.04720/) (2026)
- [VLM2Vec: Training Vision-Language Models for Massive Multimodal Embedding Tasks](https://mlsanity.com/paper/2410.05160/) (2024)
- [Visual Storytelling](https://mlsanity.com/paper/1604.03968/) (2016)
- [ReLLa: Retrieval-enhanced Large Language Models for Lifelong Sequential Behavior Comprehension in Recommendation](https://mlsanity.com/paper/2308.11131/) (2023)
- [What Makes A Good Story? Designing Composite Rewards for Visual Storytelling](https://mlsanity.com/paper/1909.05316/) (2019)
- [RzenEmbed: Towards Comprehensive Multimodal Retrieval](https://mlsanity.com/paper/2510.27350/) (2025)
- [Improve Multi-Modal Embedding Learning via Explicit Hard Negative Gradient Amplifying](https://mlsanity.com/paper/2506.02020/) (2025)
- [An Automatic Graph Construction Framework based on Large Language Models for Recommendation](https://mlsanity.com/paper/2412.18241/) (2024)
- [MM-R1: Unleashing the Power of Unified Multimodal Large Language Models for Personalized Image Generation](https://mlsanity.com/paper/2508.11433/) (2025)
- [DeepImageSearch: Benchmarking Multimodal Agents for Context-Aware Image Retrieval in Visual Histories](https://mlsanity.com/paper/2602.10809/) (2026)
- [VisChainBench: A Benchmark for Multi-Turn, Multi-Image Visual Reasoning Beyond Language Priors](https://mlsanity.com/paper/2512.06759/) (2025)

## Built on by (most cited first, 1 of 1)

- [AMBER: Multi-View Adaptive Budget Allocation for Listwise Vision-Language Reranking](https://mlsanity.com/paper/2610.02831/) (2026)

## Abstract (arXiv)

Image retrieval has traditionally been formulated as a point-wise matching problem, where each candidate image is scored in isolation. However, this atomic paradigm fails to capture the complexity of human search intent within personal photo collections, where users often seek compact visual stories bound by structural relations rather than isolated snapshots. To address this limitation, we introduce **Image Bundle Composition (IBC)**, a novel paradigm that shifts the objective from ranking individual images to dynamically composing cohesive image bundles from a massive, unstructured photo pool. Since target bundles are not predefined, IBC presents a severe combinatorial explosion challenge and demands modeling non-decomposable joint relevance. To establish this paradigm, we construct **IBCBench**, the first IBC benchmark dataset containing 109,467 images and 667 verified queries, built via a semi-automated verification pipeline. Furthermore, we propose **BundleWeaver**, an agentic framework that reformulates IBC as query-conditioned incremental hyperedge discovery. By employing a Large Language Model to adaptively search for missing relational roles and utilizing a Vision-Language Model for whole-bundle verification, BundleWeaver effectively navigates the combinatorial space. Extensive experiments demonstrate that while state-of-the-art embedding models and static decompose-and-rerank paradigms suffer from relational blindness, BundleWeaver achieves substantial performance gains, highlighting the necessity of shifting from atomic scoring to dynamic relational composition. Our dataset and code are available.