pith. sign in

arxiv: 2606.04604 · v1 · pith:RD5O3KLInew · submitted 2026-06-03 · 💻 cs.CV

COMBINER: Composed Image Retrieval Guided by Attribute-based Neighbor Relations

Zixu Li , Yupeng Hu , Zhiwei Chen , Haokun Wen , Xuemeng Song , Liqiang Nie This is my paper

Pith reviewed 2026-06-28 06:40 UTC · model grok-4.3

classification 💻 cs.CV
keywords composed image retrievalattribute prototypessemantic disentanglementneighbor relationscross-modal compositionmultimodal retrievalimage retrieval
0
0 comments X

The pith

COMBINER improves composed image retrieval by using attribute prototypes to address visually similar but attribute-unrelated samples.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces COMBINER for composed image retrieval, targeting cases where images appear visually alike yet differ in attributes. It proposes three modules to disentangle attribute features from multimodal inputs, construct unified cross-modal prototypes for composition, and model both pairwise and neighbor relations via an attribute prototype-based similarity metric. This setup claims to resolve entanglement in attribute semantics, cross-modal inconsistency, and lack of supervision signals. The result is a more accurate capture of semantic relations among samples. Experiments on three benchmark datasets support better retrieval performance than prior methods.

Core claim

COMBINER represents the first study addressing visually similar but attribute-unrelated samples in composed image retrieval. It achieves this by an attribute prototype-based similarity metric that mines dual relations, implemented through Adaptive Semantic Disentanglement for separating attribute features, Unified Prototype-based Composition for building cross-modal unified prototypes, and Dual Relations Modeling for capturing attribute-based pairwise and neighbor relations.

What carries the argument

Attribute prototype-based similarity metric in the Dual Relations Modeling module, which distinguishes samples by attribute similarity rather than visual appearance alone.

Load-bearing premise

The three core issues of attribute entanglement, modality inconsistency, and missing supervision can be resolved by the three modules without external supervision or new inconsistencies.

What would settle it

A test set of image pairs that are visually similar but differ in attributes, where retrieval accuracy does not exceed that of baseline methods using standard visual or text similarity.

Figures

Figures reproduced from arXiv: 2606.04604 by Haokun Wen, Liqiang Nie, Xuemeng Song, Yupeng Hu, Zhiwei Chen, Zixu Li.

Figure 1
Figure 1. Figure 1: Example of (a) Pairwise Relations, (b) Neighbor Rela [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Schematic of our proposed similarity measure method [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Overall framework of COMBINER, which consists of (a) Adaptive Semantic Disentanglement, (b) Unified Prototype [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: Influence of (a) Attribute Prototype Number [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: Similarity Matrix Visualization on FashionIQ. [PITH_FULL_IMAGE:figures/full_fig_p013_7.png] view at source ↗
Figure 6
Figure 6. Figure 6: Case study on (a) FashionIQ, (b) Shoes, (c) CIRR, [PITH_FULL_IMAGE:figures/full_fig_p013_6.png] view at source ↗
Figure 8
Figure 8. Figure 8: Attention Visualizations on (a) Dresses, (b) Shirts, (c) [PITH_FULL_IMAGE:figures/full_fig_p013_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Visualization of Semantic Cluster Neighbors on (a) [PITH_FULL_IMAGE:figures/full_fig_p014_9.png] view at source ↗
read the original abstract

Composed Image Retrieval (CIR) represents a challenging retrieval task that targets locating specific images through multimodal inputs. Despite recent progress in CIR techniques, prior approaches often overlook cases where images appear visually alike yet differ in attributes, potentially undermining both multimodal feature fusion and similarity modeling. To mitigate this limitation, we design a unified representation of cross-modal features based on attribute prototypes. Nevertheless, the task is far from straightforward, owing to three core issues: (1) entanglement in attribute-level semantics, (2) inconsistency across modalities, and (3) supervised signal missing. To tackle the above obstacles, we introduce a COMposed image retrieval network guided By attrIbute-based NEighbor Relations (COMBINER). Specifically, we first design an Adaptive Semantic Disentanglement module, which is capable of disentangling attribute features based on multimodal primitive features. Secondly, we propose a Unified Prototype-based Composition module, which can construct cross-modal unified prototypes (CUP) and facilitate multimodal feature composition. Finally, we introduce a Dual Relations Modeling module, which can mine pairwise and neighbor relations based on attribute similarity. Compared to traditional neighbor relations modeling CIR methods, COMBINER represents the first study addressing the phenomenon of visually similar but attribute-unrelated samples. It achieves a more accurate understanding of the semantic relations among samples by employing an attribute prototype-based similarity metric. Comprehensive experiments conducted on three benchmark datasets confirm the effectiveness of our proposed COMBINER. The implementation of our method will be accessed at https://github.com/Lee-zixu/COMBINER

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 1 minor

Summary. The manuscript proposes COMBINER, a composed image retrieval (CIR) network that targets the phenomenon of visually similar but attribute-unrelated samples. It introduces an attribute-prototype representation and three modules—Adaptive Semantic Disentanglement, Unified Prototype-based Composition (constructing cross-modal unified prototypes), and Dual Relations Modeling (mining pairwise and neighbor relations via attribute similarity)—to address attribute-level entanglement, cross-modal inconsistency, and missing supervision signals. The work claims to be the first to explicitly handle this phenomenon via an attribute prototype-based similarity metric and reports effectiveness on three benchmark datasets, with code to be released.

Significance. If the modules deliver disentanglement and unified prototypes that improve semantic relation modeling without new inconsistencies, the approach could advance CIR by providing a more accurate handling of attribute differences in visually similar images. The explicit code release is a strength for reproducibility.

major comments (3)
  1. [Abstract, §3] Abstract and §3: The central claim that the three modules jointly resolve the three core issues (entanglement, inconsistency, missing signals) without introducing new modality conflicts or implicit supervision is load-bearing for the 'first study' assertion, yet the manuscript provides only high-level module descriptions with no equations, loss terms, or architectural constraints shown to guarantee the promised disentanglement and unified prototypes (CUP).
  2. [§3.3] §3.3 (Dual Relations Modeling): The attribute prototype-based similarity metric is presented as enabling more accurate neighbor relations than traditional methods, but without the explicit definition or derivation of how this metric differs from the attribute prototypes themselves, it is unclear whether reported gains reduce to the prototype construction rather than new relational modeling.
  3. [Experiments] Experiments section: The claim of effectiveness on three benchmarks is asserted, but without ablations isolating each module's contribution to the three core issues (or controls for whether the modules introduce cross-modal inconsistencies), the support for the central performance claims remains incomplete.
minor comments (1)
  1. [Abstract] The abstract states 'The implementation of our method will be accessed at https://github.com/Lee-zixu/COMBINER' but the manuscript should include a direct link or DOI in the camera-ready version.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments, which highlight areas where additional technical detail and experimental rigor will strengthen the manuscript. We address each major comment below and commit to revisions that provide the requested equations, derivations, and ablations while preserving the core contributions.

read point-by-point responses

  1. Referee: Abstract and §3: The central claim that the three modules jointly resolve the three core issues (entanglement, inconsistency, missing signals) without introducing new modality conflicts or implicit supervision is load-bearing for the 'first study' assertion, yet the manuscript provides only high-level module descriptions with no equations, loss terms, or architectural constraints shown to guarantee the promised disentanglement and unified prototypes (CUP).

    Authors: We agree that the current high-level descriptions are insufficient to fully substantiate the joint resolution of the three issues. In the revised manuscript we will insert the full mathematical formulations for Adaptive Semantic Disentanglement (including the disentanglement loss and attribute-level constraints), the construction of cross-modal unified prototypes (CUP) with its composition equations, and the overall training objective. We will also add explicit architectural constraints and a short analysis showing that the modules do not introduce new cross-modal inconsistencies or rely on implicit supervision beyond the provided attribute labels. revision: yes

  2. Referee: §3.3 (Dual Relations Modeling): The attribute prototype-based similarity metric is presented as enabling more accurate neighbor relations than traditional methods, but without the explicit definition or derivation of how this metric differs from the attribute prototypes themselves, it is unclear whether reported gains reduce to the prototype construction rather than new relational modeling.

    Authors: We will revise §3.3 to include the precise definition of the attribute prototype-based similarity metric, its derivation from the unified prototypes, and a clear separation between the prototype construction step and the subsequent pairwise/neighbor relation modeling. This will demonstrate that the metric incorporates both attribute similarity and neighbor structure in a manner distinct from the prototypes used for composition alone. revision: yes

  3. Referee: Experiments section: The claim of effectiveness on three benchmarks is asserted, but without ablations isolating each module's contribution to the three core issues (or controls for whether the modules introduce cross-modal inconsistencies), the support for the central performance claims remains incomplete.

    Authors: We acknowledge that the current experiments lack module-specific ablations tied directly to the three core issues and explicit checks for introduced inconsistencies. The revised manuscript will add targeted ablation tables that measure each module's impact on attribute disentanglement, cross-modal consistency, and supervision signal quality, together with controls (e.g., modality-wise retrieval gaps and consistency regularization metrics) to verify that no new cross-modal conflicts are introduced. revision: yes

Circularity Check

0 steps flagged

No significant circularity; proposal of new modules remains independent of its inputs

full rationale

The abstract states three core issues and introduces three named modules (Adaptive Semantic Disentanglement, Unified Prototype-based Composition, Dual Relations Modeling) to address them, plus an attribute-prototype similarity metric. No equations, loss functions, or derivation steps are supplied that would allow any claimed performance gain or 'first study' status to reduce by construction to the module definitions themselves. No self-citations, fitted parameters renamed as predictions, or ansatzes smuggled via prior work appear. The central claim is therefore a standard architectural proposal whose correctness must be judged by external benchmarks rather than internal definitional equivalence.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 2 invented entities

The central claim rests on the introduction of attribute prototypes and the three modules as solutions to the three core issues; no free parameters, standard mathematical axioms, or independently evidenced entities are described in the abstract.

invented entities (2)
  • attribute prototypes no independent evidence
    purpose: Unified representation of cross-modal attribute-level semantics to enable disentanglement and similarity measurement
    Introduced as the foundational construct for handling the three core issues; no independent evidence supplied in abstract.
  • cross-modal unified prototypes (CUP) no independent evidence
    purpose: To construct consistent representations across image and text modalities for feature composition
    New construct proposed in the Unified Prototype-based Composition module; no external validation shown.

pith-pipeline@v0.9.1-grok · 5825 in / 1355 out tokens · 32546 ms · 2026-06-28T06:40:42.287336+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

99 extracted references · 9 linked inside Pith

  1. [1]

    Tempret: Temporal enhancement and two- stage reranking for cvpr 2026 epic-kitchens-100 multi-instance retrieval challenge.arXiv preprint arXiv:2605.24470, 2026

    Zixu Li, Yupeng Hu, Zhiwei Chen, Zhiheng Fu, Xiaowei Zhu, Weili Guan, and Liqiang Nie. Tempret: Temporal enhancement and two- stage reranking for cvpr 2026 epic-kitchens-100 multi-instance retrieval challenge.arXiv preprint arXiv:2605.24470, 2026

    Pith/arXiv arXiv 2026

  2. [2]

    Stable: Efficient hybrid nearest neighbor search via magnitude-uniformity and cardinality-robustness.IEEE TKDE, 2026

    Qianyun Yang, Zhiwei Chen, Yupeng Hu, Zixu Li, Zhiheng Fu, and Liqiang Nie. Stable: Efficient hybrid nearest neighbor search via magnitude-uniformity and cardinality-robustness.IEEE TKDE, 2026

    2026

  3. [3]

    Erase: Bypassing collaborative detection of ai counterfeit via comprehensive artifacts elimination.IEEE TDSC, pages 1–18, Mar

    Qianyun Yang, Peizhuo Lv, Yingjiu Li, Shengzhi Zhang, Yuxuan Chen, Zhiwei Chen, Zixu Li, and Yupeng Hu. Erase: Bypassing collaborative detection of ai counterfeit via comprehensive artifacts elimination.IEEE TDSC, pages 1–18, Mar. 2026

    2026

  4. [4]

    User: Unified semantic enhancement with momentum contrast for image-text retrieval.IEEE Transactions on Image Processing, 33:595–609, 2024

    Yan Zhang, Zhong Ji, Di Wang, Yanwei Pang, and Xuelong Li. User: Unified semantic enhancement with momentum contrast for image-text retrieval.IEEE Transactions on Image Processing, 33:595–609, 2024

    2024

  5. [5]

    Deep boosting learning: a brand-new cooperative approach for image- text matching.IEEE Transactions on Image Processing, 2024

    Haiwen Diao, Ying Zhang, Shang Gao, Xiang Ruan, and Huchuan Lu. Deep boosting learning: a brand-new cooperative approach for image- text matching.IEEE Transactions on Image Processing, 2024

    2024

  6. [6]

    Decoupled cross-modal phrase-attention network for image- sentence matching.IEEE Transactions on Image Processing, 33:1326– 1337, 2022

    Zhangxiang Shi, Tianzhu Zhang, Xi Wei, Feng Wu, and Yongdong Zhang. Decoupled cross-modal phrase-attention network for image- sentence matching.IEEE Transactions on Image Processing, 33:1326– 1337, 2022

    2022

  7. [7]

    Semantics disentangling for cross-modal retrieval.IEEE Trans- actions on Image Processing, 33:2226–2237, 2024

    Zheng Wang, Xing Xu, Jiwei Wei, Ning Xie, Yang Yang, and Heng Tao Shen. Semantics disentangling for cross-modal retrieval.IEEE Trans- actions on Image Processing, 33:2226–2237, 2024

    2024

  8. [8]

    Refine: Composed video retrieval via shared and differential semantics enhancement.ACM ToMM, 2026

    Yupeng Hu, Zixu Li, Zhiwei Chen, Qinlei Huang, Zhiheng Fu, Mingzhu Xu, and Liqiang Nie. Refine: Composed video retrieval via shared and differential semantics enhancement.ACM ToMM, 2026

    2026

  9. [9]

    Composing text and image for image retrieval - an empirical odyssey

    Nam V o, Lu Jiang, Chen Sun, Kevin Murphy, Li-Jia Li, Li Fei-Fei, and James Hays. Composing text and image for image retrieval - an empirical odyssey. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 6439–6448. IEEE, 2019

    2019

  10. [10]

    Hint: Composed image retrieval with dual-path compositional contextualized network

    Mingyu Zhang, Zixu Li, Zhiwei Chen, Zhiheng Fu, Xiaowei Zhu, Jiajia Nie, Yinwei Wei, and Yupeng Hu. Hint: Composed image retrieval with dual-path compositional contextualized network. InICASSP, pages 13002–13006. IEEE, 2026

    2026

  11. [11]

    Melt: Improve composed image retrieval via the modification frequentation-rarity balance network

    Guozhi Qiu, Zhiwei Chen, Zixu Li, Qinlei Huang, Zhiheng Fu, Xuemeng Song, and Yupeng Hu. Melt: Improve composed image retrieval via the modification frequentation-rarity balance network. InICASSP, pages 13007–13011. IEEE, 2026

    2026

  12. [12]

    Air-know: Arbiter-calibrated knowledge-internalizing robust network for composed image retrieval.arXiv preprint arXiv:2604.19386, 2026

    Zhiheng Fu, Yupeng Hu, Qianyun Yang, Shiqi Zhang, Zhiwei Chen, and Zixu Li. Air-know: Arbiter-calibrated knowledge-internalizing robust network for composed image retrieval.arXiv preprint arXiv:2604.19386, 2026

    Pith/arXiv arXiv 2026

  13. [13]

    Conesep: Cone-based robust noise-unlearning com- positional network for composed image retrieval.arXiv preprint arXiv:2604.20358, 2026

    Zixu Li, Yupeng Hu, Zhiwei Chen, Mingyu Zhang, Zhiheng Fu, and Liqiang Nie. Conesep: Cone-based robust noise-unlearning com- positional network for composed image retrieval.arXiv preprint arXiv:2604.20358, 2026

    Pith/arXiv arXiv 2026

  14. [14]

    Mmerror: A benchmark for erroneous reasoning in vision-language models.arXiv preprint arXiv:2601.03331, 2026

    Yang Shi, Yifeng Xie, Minzhe Guo, Liangsi Lu, Mingxuan Huang, Jingchao Wang, Zhihong Zhu, Boyan Xu, and Zhiqi Huang. Mmerror: A benchmark for erroneous reasoning in vision-language models.arXiv preprint arXiv:2601.03331, 2026

    Pith/arXiv arXiv 2026

  15. [15]

    Egoaction: Egocentric action composition with reliability-aware temporal fusion for the epic-kitchens action detection challenge at cvpr 2026.arXiv preprint arXiv:2605.24496, 2026

    Zhiheng Fu, Zixu Li, Zhiwei Chen, Fangxu Liu, Yupeng Hu, Weili Guan, and Liqiang Nie. Egoaction: Egocentric action composition with reliability-aware temporal fusion for the epic-kitchens action detection challenge at cvpr 2026.arXiv preprint arXiv:2605.24496, 2026

    Pith/arXiv arXiv 2026

  16. [16]

    Detecting congestion-related attacks via fine-grained queue diagnosis

    Rui Dai, Dan Tang, Zheng Qin, Kai Chen, Keqin Li, and Jiliang Zhang. Detecting congestion-related attacks via fine-grained queue diagnosis. IEEE Transactions on Cognitive Communications and Networking, 2025

    2025

  17. [17]

    Mlp-slam: Multilayer perceptron-based simul- taneous localization and mapping.arXiv preprint arXiv:2410.10669, 2024

    Taozhe Li and Wei Sun. Mlp-slam: Multilayer perceptron-based simul- taneous localization and mapping.arXiv preprint arXiv:2410.10669, 2024

    arXiv 2024

  18. [18]

    Mwd-cfm: Detection and mitigation of ddos attack against sdn flow tables.IEEE Transactions on Networking, 34:4269–4282, 2026

    Dan Tang, Chenguang Zuo, Xinmeng Li, Siyuan Wang, Wei Liang, Keqin Li, and Jiliang Zhang. Mwd-cfm: Detection and mitigation of ddos attack against sdn flow tables.IEEE Transactions on Networking, 34:4269–4282, 2026

    2026

  19. [19]

    Event-triggered adaptive tracking control for usv based on enhanced optimized backstepping technique.ISA transactions, 2025

    Hugan Zhang, Xianku Zhang, Yongjin Liu, Shihang Gao, and Daocheng Ma. Event-triggered adaptive tracking control for usv based on enhanced optimized backstepping technique.ISA transactions, 2025

    2025

  20. [20]

    Egoadapt: A multi-scene egocentric adaptation method for cvpr 2026 hd-epic vqa challenge.arXiv preprint arXiv:2605.24500, 2026

    Zhiwei Chen, Yupeng Hu, Zixu Li, Zhiheng Fu, Guozhi Qiu, Weili Guan, and Liqiang Nie. Egoadapt: A multi-scene egocentric adaptation method for cvpr 2026 hd-epic vqa challenge.arXiv preprint arXiv:2605.24500, 2026

    Pith/arXiv arXiv 2026

  21. [21]

    Omniego-r 2: A routed reasoning framework for the 1st cross-domain egocross challenge at cvpr 2026.arXiv preprint arXiv:2605.24481, 2026

    Zixu Li, Zhiwei Chen, Zhiheng Fu, Wenbo Wang, Yupeng Hu, Weili Guan, and Liqiang Nie. Omniego-r 2: A routed reasoning framework for the 1st cross-domain egocross challenge at cvpr 2026.arXiv preprint arXiv:2605.24481, 2026

    Pith/arXiv arXiv 2026

  22. [22]

    R 3: Composed video retrieval via reasoning-guided recalling and re-ranking.arXiv preprint arXiv:2606.01113, 2026

    Zixu Li, Yupeng Hu, Zhiheng Fu, Zhiwei Chen, Weili Guan, and Liqiang Nie. R 3: Composed video retrieval via reasoning-guided recalling and re-ranking.arXiv preprint arXiv:2606.01113, 2026

    Pith/arXiv arXiv 2026

  23. [23]

    Core-mmrag: Cross-source knowledge reconciliation for multimodal rag

    Yang Tian, Fan Liu, Jingyuan Zhang, Yupeng Hu, Liqiang Nie, et al. Core-mmrag: Cross-source knowledge reconciliation for multimodal rag. InProceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 32967– 32982, 2025

    2025

  24. [24]

    Chordedit: One-step low-energy transport for image editing

    Liangsi Lu, Xuhang Chen, Minzhe Guo, Shichu Li, Jingchao Wang, and Yang Shi. Chordedit: One-step low-energy transport for image editing. arXiv preprint arXiv:2602.19083, 2026

    arXiv 2026

  25. [25]

    Semantic collaborative learning for cross-modal moment localization

    Yupeng Hu, Kun Wang, Meng Liu, Haoyu Tang, and Liqiang Nie. Semantic collaborative learning for cross-modal moment localization. 15 ACM Transactions on Information Systems, 42(2):1–26, 2023

    2023

  26. [26]

    Infor- mation guided levy flight for robot search in unknown environments

    Weitao Zhao, Zati Hakim Azizul, Xin Lyu, and Weijie Kuang. Infor- mation guided levy flight for robot search in unknown environments. Journal of King Saud University Computer and Information Sciences, 2026

    2026

  27. [27]

    Coarse-to-fine semantic alignment for cross-modal moment localization.IEEE Transactions on Image Processing, 30:5933– 5943, 2021

    Yupeng Hu, Liqiang Nie, Meng Liu, Kun Wang, Yinglong Wang, and Xian-Sheng Hua. Coarse-to-fine semantic alignment for cross-modal moment localization.IEEE Transactions on Image Processing, 30:5933– 5943, 2021

    2021

  28. [28]

    Grain: Gravity-resistance adaptive framework for identifying influential nodes using multi-order structural diversity.Information Processing & Management, 63(4):104618, 2026

    Yirun Ruan, Xinghua Qin, Sizheng Liu, Mengmeng Zhang, Jun Tang, Yanming Guo, and Tianyuan Yu. Grain: Gravity-resistance adaptive framework for identifying influential nodes using multi-order structural diversity.Information Processing & Management, 63(4):104618, 2026

    2026

  29. [29]

    Angel or devil: Discriminating hard samples and anomaly contaminations for unsupervised time series anomaly detection.Neural Networks, page 108532, 2026

    Ruyi Zhang, Hongzuo Xu, Songlei Jian, Yusong Tan, Haifang Zhou, and Rulin Xu. Angel or devil: Discriminating hard samples and anomaly contaminations for unsupervised time series anomaly detection.Neural Networks, page 108532, 2026

    2026

  30. [30]

    Video moment localization via deep cross-modal hashing.IEEE Transactions on Image Processing, 30:4667–4677, 2021

    Yupeng Hu, Meng Liu, Xiaobin Su, Zan Gao, and Liqiang Nie. Video moment localization via deep cross-modal hashing.IEEE Transactions on Image Processing, 30:4667–4677, 2021

    2021

  31. [31]

    Progressive learning for image retrieval with hybrid-modality queries

    Yida Zhao, Yuqing Song, and Qin Jin. Progressive learning for image retrieval with hybrid-modality queries. InProceedings of the International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 1012–1021, 2022

    2022

  32. [32]

    Sentence-level prompts benefit composed image retrieval

    Xinxing Xu, Yong Liu, Salman Khan, Fahad Khan, Wangmeng Zuo, Rick Siow Mong Goh, Chun-Mei Feng, et al. Sentence-level prompts benefit composed image retrieval. InInternational Conference on Learning Representations, 2024

    2024

  33. [33]

    Decomposing semantic shifts for composed image re- trieval

    Xingyu Yang, Daqing Liu, Heng Zhang, Yong Luo, Chaoyue Wang, and Jing Zhang. Decomposing semantic shifts for composed image re- trieval. InProceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 6576–6584, 2024

    2024

  34. [34]

    Relieving triplet ambiguity: Consensus network for language-guided image re- trieval.arXiv preprint arXiv:2306.02092, 2023

    Xu Zhang, Zhedong Zheng, Xiaohan Wang, and Yi Yang. Relieving triplet ambiguity: Consensus network for language-guided image re- trieval.arXiv preprint arXiv:2306.02092, 2023

    arXiv 2023

  35. [35]

    Ranking-aware uncertainty for text- guided image retrieval.arXiv preprint arXiv:2308.08131, 2023

    Junyang Chen and Hanjiang Lai. Ranking-aware uncertainty for text- guided image retrieval.arXiv preprint arXiv:2308.08131, 2023

    arXiv 2023

  36. [36]

    Target-guided composed image retrieval

    Haokun Wen, Xian Zhang, Xuemeng Song, Yinwei Wei, and Liqiang Nie. Target-guided composed image retrieval. InProceedings of the ACM International Conference on Multimedia, pages 915–923, 2023

    2023

  37. [37]

    Semantic distil- lation from neighborhood for composed image retrieval

    Yifan Wang, Wuliang Huang, Lei Li, and Chun Yuan. Semantic distil- lation from neighborhood for composed image retrieval. InProceedings of the ACM International Conference on Multimedia, 2024

    2024

  38. [38]

    Learn- ing attribute-driven disentangled representations for interactive fashion retrieval

    Yuxin Hou, Eleonora Vig, Michael Donoser, and Loris Bazzani. Learn- ing attribute-driven disentangled representations for interactive fashion retrieval. InProceedings of the IEEE/CVF International conference on computer vision, pages 12147–12157, 2021

    2021

  39. [39]

    Face image retrieval with attribute manipulation

    Alireza Zaeemzadeh, Shabnam Ghadar, Baldo Faieta, Zhe Lin, Nazanin Rahnavard, Mubarak Shah, and Ratheesh Kalarot. Face image retrieval with attribute manipulation. InProceedings of the IEEE/CVF interna- tional conference on computer vision, pages 12116–12125, 2021

    2021

  40. [40]

    Generative attribute manipulation scheme for flexible fashion search

    Xin Yang, Xuemeng Song, Xianjing Han, Haokun Wen, Jie Nie, and Liqiang Nie. Generative attribute manipulation scheme for flexible fashion search. InProceedings of the 43rd international acm sigir conference on research and development in information retrieval, pages 941–950, 2020

    2020

  41. [41]

    Composed image retrieval via cross relation network with hierarchical aggregation transformer.IEEE Transactions on Image Processing, 2023

    Qu Yang, Mang Ye, Zhaohui Cai, Kehua Su, and Bo Du. Composed image retrieval via cross relation network with hierarchical aggregation transformer.IEEE Transactions on Image Processing, 2023

    2023

  42. [42]

    Composed image retrieval via explicit erasure and replenishment with semantic alignment.IEEE Transactions on Image Processing, 31:5976– 5988, 2022

    Gangjian Zhang, Shikui Wei, Huaxin Pang, Shuang Qiu, and Yao Zhao. Composed image retrieval via explicit erasure and replenishment with semantic alignment.IEEE Transactions on Image Processing, 31:5976– 5988, 2022

    2022

  43. [43]

    Multimodal composition example mining for composed query image retrieval.IEEE Transactions on Image Processing, 33:1149–1161, 2024

    Gangjian Zhang, Shikun Li, Shikui Wei, Shiming Ge, Na Cai, and Yao Zhao. Multimodal composition example mining for composed query image retrieval.IEEE Transactions on Image Processing, 33:1149–1161, 2024

    2024

  44. [44]

    Finecir: Explicit parsing of fine-grained modification se- mantics for composed image retrieval.https://arxiv.org/abs/2503.21309, 2025

    Zixu Li, Zhiheng Fu, Yupeng Hu, Zhiwei Chen, Haokun Wen, and Liqiang Nie. Finecir: Explicit parsing of fine-grained modification se- mantics for composed image retrieval.https://arxiv.org/abs/2503.21309, 2025

    arXiv 2025

  45. [45]

    Pair: Complementarity-guided disentan- glement for composed image retrieval

    Zhiheng Fu, Zixu Li, Zhiwei Chen, Chunxiao Wang, Xuemeng Song, Yupeng Hu, and Liqiang Nie. Pair: Complementarity-guided disentan- glement for composed image retrieval. InProceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing, pages 1–5. IEEE, 2025

    2025

  46. [46]

    Median: Adaptive intermediate-grained aggregation network for composed image retrieval

    Qinlei Huang, Zhiwei Chen, Zixu Li, Chunxiao Wang, Xuemeng Song, Yupeng Hu, and Liqiang Nie. Median: Adaptive intermediate-grained aggregation network for composed image retrieval. InProceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing, pages 1–5. IEEE, 2025

    2025

  47. [47]

    Candi- date set re-ranking for composed image retrieval with dual multi-modal encoder.Transactions on Machine Learning Research, 2024

    Zheyuan Liu, Weixuan Sun, Damien Teney, and Stephen Gould. Candi- date set re-ranking for composed image retrieval with dual multi-modal encoder.Transactions on Machine Learning Research, 2024

    2024

  48. [48]

    Simple but effective raw-data level multimodal fusion for composed image retrieval

    Haokun Wen, Xuemeng Song, Xiaolin Chen, Yinwei Wei, Liqiang Nie, and Tat-Seng Chua. Simple but effective raw-data level multimodal fusion for composed image retrieval. InProceedings of the International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 229–239, 2024

    2024

  49. [49]

    Language-only training of zero-shot composed image retrieval

    Geonmo Gu, Sanghyuk Chun, Wonjae Kim, , Yoohoon Kang, and Sangdoo Yun. Language-only training of zero-shot composed image retrieval. InConference on Computer Vision and Pattern Recognition, 2024

    2024

  50. [50]

    Semantic editing increment benefits zero-shot composed image retrieval

    Zhenyu Yang, Shengsheng Qian, Dizhan Xue, Jiahong Wu, Fan Yang, Weiming Dong, and Changsheng Xu. Semantic editing increment benefits zero-shot composed image retrieval. InProceedings of the ACM International Conference on Multimedia, pages 1245–1254, 2024

    2024

  51. [51]

    MagicLens: Self-supervised image retrieval with open-ended instructions

    Kai Zhang, Yi Luan, Hexiang Hu, Kenton Lee, Siyuan Qiao, Wenhu Chen, Yu Su, and Ming-Wei Chang. MagicLens: Self-supervised image retrieval with open-ended instructions. InProceedings of the International Conference on Machine Learning, pages 59403–59420, 2024

    2024

  52. [52]

    Offset: Segmentation-based focus shift revision for composed image retrieval

    Zhiwei Chen, Yupeng Hu, Zixu Li, Zhiheng Fu, Xuemeng Song, and Liqiang Nie. Offset: Segmentation-based focus shift revision for composed image retrieval. InProceedings of the ACM International Conference on Multimedia, page 61136122, 2025

    2025

  53. [53]

    Hud: Hierarchical uncertainty-aware disambiguation network for composed video retrieval

    Zhiwei Chen, Yupeng Hu, Zixu Li, Zhiheng Fu, Haokun Wen, and Weili Guan. Hud: Hierarchical uncertainty-aware disambiguation network for composed video retrieval. InProceedings of the ACM International Conference on Multimedia, page 61436152, 2025

    2025

  54. [54]

    Composed image retrieval with text feedback via multi-grained uncertainty regularization

    Yiyang Chen, Zhedong Zheng, Wei Ji, Leigang Qu, and Tat-Seng Chua. Composed image retrieval with text feedback via multi-grained uncertainty regularization. InInternational Conference on Learning Representations, 2024

    2024

  55. [55]

    Cosmo: Content- style modulation for image retrieval with text feedback

    Seungmin Lee, Dongwan Kim, and Bohyung Han. Cosmo: Content- style modulation for image retrieval with text feedback. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 802–812. IEEE, 2021

    2021

  56. [56]

    Comprehensive linguistic-visual composition network for image retrieval

    Haokun Wen, Xuemeng Song, Xin Yang, Yibing Zhan, and Liqiang Nie. Comprehensive linguistic-visual composition network for image retrieval. InProceedings of the International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 1369–

  57. [57]

    Self-training boosted multi-factor matching network for composed image retrieval.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023

    Haokun Wen, Xuemeng Song, Jianhua Yin, Jianlong Wu, Weili Guan, and Liqiang Nie. Self-training boosted multi-factor matching network for composed image retrieval.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023

    2023

  58. [58]

    Tema: Anchor the image, follow the text for multi-modification composed image retrieval.arXiv preprint arXiv:2604.21806, 2026

    Zixu Li, Yupeng Hu, Zhiheng Fu, Zhiwei Chen, Yongqi Li, and Liqiang Nie. Tema: Anchor the image, follow the text for multi-modification composed image retrieval.arXiv preprint arXiv:2604.21806, 2026

    Pith/arXiv arXiv 2026

  59. [59]

    Habit: Chrono-synergia robust progressive learning framework for composed image retrieval

    Zixu Li, Yupeng Hu, Zhiwei Chen, Shiqi Zhang, Qinlei Huang, Zhiheng Fu, and Yinwei Wei. Habit: Chrono-synergia robust progressive learning framework for composed image retrieval. InAAAI, volume 40, pages 6762–6770, 2026

    2026

  60. [60]

    Set of diverse queries with uncertainty regularization for composed image retrieval.IEEE Transactions on Circuits and Systems for Video Technology, 2024

    Yahui Xu, Jiwei Wei, Yi Bin, Yang Yang, Zeyu Ma, and Heng Tao Shen. Set of diverse queries with uncertainty regularization for composed image retrieval.IEEE Transactions on Circuits and Systems for Video Technology, 2024

    2024

  61. [61]

    Intent: Invariance and discrimination-aware noise mitigation for robust composed image retrieval

    Zhiwei Chen, Yupeng Hu, Zhiheng Fu, Zixu Li, Jiale Huang, Qinlei Huang, and Yinwei Wei. Intent: Invariance and discrimination-aware noise mitigation for robust composed image retrieval. InAAAI, vol- ume 40, pages 20463–20471, 2026

    2026

  62. [62]

    Retrack: Evidence-driven dual-stream directional anchor calibration network for composed video retrieval

    Zixu Li, Yupeng Hu, Zhiwei Chen, Qinlei Huang, Guozhi Qiu, Zhiheng Fu, and Meng Liu. Retrack: Evidence-driven dual-stream directional anchor calibration network for composed video retrieval. InAAAI, volume 40, pages 23373–23381, 2026

    2026

  63. [63]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. InInternational conference on machine learning, pages 8748–8763. PMLR, 2021

    2021

  64. [64]

    Effective conditioned and composed image retrieval com- bining clip-based features

    Alberto Baldrati, Marco Bertini, Tiberio Uricchio, and Alberto Del Bimbo. Effective conditioned and composed image retrieval com- bining clip-based features. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 21466–21474, 2022

    2022

  65. [65]

    High reliability multi-input converter with low input current ripple based on sepic for solar-powered unmanned aerial vehicle.IEEE Transactions on 16 Consumer Electronics, 2026

    Binxin Zhu, Wenxin Liao, Xiaoli She, and Jinhai An. High reliability multi-input converter with low input current ripple based on sepic for solar-powered unmanned aerial vehicle.IEEE Transactions on 16 Consumer Electronics, 2026

    2026

  66. [66]

    Training-free multi- style fusion through reference-based adaptive modulation, 2025

    Xu Liu, Yibo Lu, Xinxian Wang, and Xinyu Wu. Training-free multi- style fusion through reference-based adaptive modulation, 2025

    2025

  67. [67]

    Don’t let the information slip away.arXiv preprint arXiv:2602.22595, 2026

    Taozhe Li. Don’t let the information slip away.arXiv preprint arXiv:2602.22595, 2026

    arXiv 2026

  68. [68]

    Dnsgreen: A comprehensive defense system against bounce-style dns ddos attacks with p4.IEEE Transactions on Computers, 2025

    Dan Tang, Xiaocai Wang, Pei Tan, Zheng Qin, Keqin Li, and Jiliang Zhang. Dnsgreen: A comprehensive defense system against bounce-style dns ddos attacks with p4.IEEE Transactions on Computers, 2025

    2025

  69. [69]

    Prompt-guided dual latent steering for inversion problems, 2025

    Yichen Wu, Xu Liu, Chenxuan Zhao, and Xinyu Wu. Prompt-guided dual latent steering for inversion problems, 2025

    2025

  70. [70]

    Machine learning-driven simulation and optimization of phosphate adsorption on metal-organic frameworks

    Jie Huang, Ziang Zong, Penghui Wang, Yuxuan Zhang, Degui Gao, Yingqi Wang, and Zhanjun Li. Machine learning-driven simulation and optimization of phosphate adsorption on metal-organic frameworks. Separation and Purification Technology, page 137479, 2026

    2026

  71. [71]

    Attribute prototype network for zero-shot learning.Advances in Neural Information Processing Systems, 33:21969–21980, 2020

    Wenjia Xu, Yongqin Xian, Jiuniu Wang, Bernt Schiele, and Zeynep Akata. Attribute prototype network for zero-shot learning.Advances in Neural Information Processing Systems, 33:21969–21980, 2020

    2020

  72. [72]

    Prototype-guided saliency feature learning for person search

    Hanjae Kim, Sunghun Joung, Ig-Jae Kim, and Kwanghoon Sohn. Prototype-guided saliency feature learning for person search. InPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4865–4874, 2021

    2021

  73. [73]

    Robust classification with convolutional prototype learning

    Hong-Ming Yang, Xu-Yao Zhang, Fei Yin, and Cheng-Lin Liu. Robust classification with convolutional prototype learning. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 3474–3482, 2018

    2018

  74. [74]

    Prototypical matching and open set rejection for zero-shot semantic segmentation

    Hui Zhang and Henghui Ding. Prototypical matching and open set rejection for zero-shot semantic segmentation. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 6974– 6983, 2021

    2021

  75. [75]

    Prototypical networks for few-shot learning.Advances in neural information processing systems, 30, 2017

    Jake Snell, Kevin Swersky, and Richard Zemel. Prototypical networks for few-shot learning.Advances in neural information processing systems, 30, 2017

    2017

  76. [76]

    Intermediate prototype mining transformer for few-shot semantic segmentation.Ad- vances in Neural Information Processing Systems, 35:38020–38031, 2022

    Yuanwei Liu, Nian Liu, Xiwen Yao, and Junwei Han. Intermediate prototype mining transformer for few-shot semantic segmentation.Ad- vances in Neural Information Processing Systems, 35:38020–38031, 2022

    2022

  77. [77]

    Interactive segmentation with prototype learning for few-shot root annotation.IEEE Transactions on Geoscience and Remote Sensing, 2025

    Xiaolei Guo, Alina Zare, Lisa Anthony, and Felix B Fritschi. Interactive segmentation with prototype learning for few-shot root annotation.IEEE Transactions on Geoscience and Remote Sensing, 2025

    2025

  78. [78]

    Rethinking semantic segmentation: A prototype view

    Tianfei Zhou, Wenguan Wang, Ender Konukoglu, and Luc Van Gool. Rethinking semantic segmentation: A prototype view. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2582–2593, 2022

    2022

  79. [79]

    Neural discrete representation learning.Advances in neural information processing systems, 30, 2017

    Aaron Van Den Oord, Oriol Vinyals, et al. Neural discrete representation learning.Advances in neural information processing systems, 30, 2017

    2017

  80. [80]

    Conditioned and composed image retrieval combining and partially fine-tuning clip-based features

    Alberto Baldrati, Marco Bertini, Tiberio Uricchio, and Alberto Del Bimbo. Conditioned and composed image retrieval combining and partially fine-tuning clip-based features. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4959–4968, 2022

    2022

Showing first 80 references.