FAST-MEL: A Fast, Accurate, and Storage Efficient Solution for Multimodal Entity Linking

Derrien Thomas; Laurent Amsaleg; Pascale S\'ebillot

arxiv: 2606.11749 · v1 · pith:S236L3L5new · submitted 2026-06-10 · 💻 cs.IR

FAST-MEL: A Fast, Accurate, and Storage Efficient Solution for Multimodal Entity Linking

Derrien Thomas , Laurent Amsaleg , Pascale S\'ebillot This is my paper

Pith reviewed 2026-06-27 08:34 UTC · model grok-4.3

classification 💻 cs.IR

keywords multimodal entity linkingknowledge basevector representationefficiencystorage optimizationencoder-based modelinformation retrieval

0 comments

The pith

FAST-MEL uses a compact fixed-size vector for each entity's text and images to link multimodal mentions at the accuracy of top systems while running three orders of magnitude faster and using one order of magnitude less storage.

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

Multimodal entity linking requires matching textual and visual mentions in data to entries in a knowledge base, but existing systems cannot deliver high accuracy, speed, and compact storage together. The paper introduces FAST-MEL, a lightweight encoder-based approach that creates a single compact vector representation for both the textual and visual aspects of each entity and mention. This design allows the system to match the linking accuracy of the strongest prior methods while executing three orders of magnitude faster and requiring ten times less index storage than the quickest alternatives. A sympathetic reader would care because the efficiency gains make large-scale multimodal linking feasible in practical applications where current solutions are too slow or too memory-heavy to deploy.

Core claim

FAST-MEL is a lightweight encoder-based MEL solution that relies on a novel and compact fixed-size vectorized representation of both the textual and visual information of each entity or mention. It matches the accuracy of the best systems but performs three orders of magnitude faster. It also consumes one order of magnitude less storage than the fastest systems.

What carries the argument

A novel compact fixed-size vectorized representation that jointly encodes textual and visual information for each entity or mention.

If this is right

MEL systems can now operate on much larger knowledge bases without prohibitive compute or memory costs.
Real-time multimodal applications such as social media monitoring or visual search become practical at scale.
The same compact representation can serve both entity linking and downstream tasks that reuse the index.
Storage savings allow deployment on resource-constrained devices or cloud instances with smaller memory footprints.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

If the vector representation generalizes, similar compact encodings could improve efficiency in other multimodal retrieval tasks such as image captioning or cross-modal search.
The speed gains suggest the method could support continuous updating of knowledge bases in streaming data scenarios.
Because storage is reduced, organizations with very large entity catalogs could maintain multiple versions of the index for different domains.

Load-bearing premise

That a compact fixed-size vector can retain enough detail from both text and images to keep linking accuracy high while delivering the claimed speed and storage savings.

What would settle it

Running the system on a large multimodal test set and observing either a drop below the accuracy of the best prior systems or failure to achieve three orders of magnitude speedup and one order of magnitude storage reduction.

Figures

Figures reproduced from arXiv: 2606.11749 by Derrien Thomas, Laurent Amsaleg, Pascale S\'ebillot.

**Figure 1.** Figure 1: Example of MEL from WikiMEL. context alongside text in order to reduce ambiguity and improve linking accuracy. A typical MEL setup is illustrated in [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗

**Figure 2.** Figure 2: Impact of the Final Embedding Dimension 𝑑 on H@1 Performance. experiments demonstrate that FAST-MEL achieves strong performance without compromising efficiency. Acknowledgments This work was in part publicly funded through the French ANR (Agence Nationale de la Recherche) under the project AGAPE with the reference ANR-24-CE38-7253. References [1] Zhiwei Hu, Víctor Gutiérrez-Basulto, Ru Li, and Jeff Z. Pan… view at source ↗

read the original abstract

Multimodal entity linking (MEL) is the task that consists of matching textual and visual mentions of entities in unstructured data to their corresponding entities in a knowledge base (KB). To be effective in large-scale practical settings, MEL systems must meet three objectives: high linking accuracy, computational efficiency, and storage efficiency, i.e., a compact yet efficient index of the KB. In this paper, we highlight that state-of-the-art systems fail to simultaneously satisfy these 3 requirements. To meet this three-fold objective, we propose FAST-MEL, a lightweight encoder-based MEL solution that relies on a novel and compact fixed-size vectorized representation of both the textual and visual information of each entity or mention. It matches the accuracy of the best systems but performs three orders of magnitude faster. It also consumes one order of magnitude less storage than the fastest systems.

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.

Desk Editor's Note private letter to a colleague

FAST-MEL introduces a compact fixed-size vector for multimodal entity linking that claims SOTA accuracy with 1000x speedup and 10x less storage, but the abstract provides no experiments or method details to check if the representation actually works.

read the letter

The main point is that the authors built a lightweight encoder-based system around a new compact fixed-size vector that packs both text and visual information for each entity and mention. They argue this hits the three goals at once—accuracy, speed, and low storage—where prior MEL work falls short on at least one.

They do a clean job laying out why existing systems cannot satisfy all three requirements simultaneously in large-scale settings. The focus on practical deployment constraints is useful for anyone who has to run MEL on real data volumes.

The soft spot is that the material here is only the abstract. No architecture details, no derivation of the vector, no datasets, no tables, and no error bars appear. The load-bearing claim is that the compact representation keeps enough multimodal signal for accurate linking; without seeing how it is built or the results, that cannot be tested. The citation pattern and direct comparisons to prior encoder methods are also absent from what is visible.

This is for people working on scalable information retrieval pipelines who need efficiency numbers they can actually use. A reader who wants to know whether the efficiency gains are real would get value only if the full paper contains reproducible experiments.

I would send it to peer review so the methods and results can be examined properly.

Referee Report

1 major / 0 minor

Summary. The paper proposes FAST-MEL, a lightweight encoder-based solution for multimodal entity linking (MEL) that relies on a novel compact fixed-size vectorized representation of textual and visual information for entities and mentions. It claims this approach simultaneously achieves high linking accuracy comparable to state-of-the-art systems, three orders of magnitude faster inference, and one order of magnitude lower storage than the fastest existing systems, addressing the failure of prior MEL systems to meet all three requirements at once.

Significance. If the performance claims hold under rigorous evaluation, the result would be significant for practical large-scale MEL deployments, as it targets the joint optimization of accuracy, speed, and storage that current systems do not achieve. The introduction of a compact multimodal representation could influence indexing strategies in knowledge-base applications.

major comments (1)

[Abstract] Abstract: the central claims of matching SOTA accuracy, 1000x speedup, and 10x storage reduction are asserted without any experimental results, tables, figures, baselines, error bars, or methodological details. The load-bearing assumption that the novel fixed-size vectorized representation preserves sufficient multimodal information for accurate linking therefore cannot be evaluated from the provided material.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their review. The comment focuses on the abstract's presentation of results. We address this below and note that the full manuscript contains the supporting experimental details.

read point-by-point responses

Referee: [Abstract] Abstract: the central claims of matching SOTA accuracy, 1000x speedup, and 10x storage reduction are asserted without any experimental results, tables, figures, baselines, error bars, or methodological details. The load-bearing assumption that the novel fixed-size vectorized representation preserves sufficient multimodal information for accurate linking therefore cannot be evaluated from the provided material.

Authors: The abstract is intended as a concise summary of the paper's main contributions and findings. The full manuscript provides the requested details: Section 3 describes the compact fixed-size vectorized representation and its construction from textual and visual features; Section 4 presents the experimental evaluation, including direct comparisons to state-of-the-art baselines with accuracy metrics, inference latency measurements (demonstrating three orders of magnitude speedup), storage footprint analysis (one order of magnitude reduction), and supporting tables and figures. Error bars and statistical details are included where appropriate in the results. These experiments directly validate that the representation preserves sufficient multimodal information for linking accuracy comparable to prior systems. revision: no

Circularity Check

0 steps flagged

No significant circularity

full rationale

The provided abstract and context contain no equations, derivations, or load-bearing self-citations. The central claim is an empirical proposal of a compact vectorized representation for MEL that achieves stated efficiency/accuracy tradeoffs; no step reduces by construction to fitted inputs, self-definitions, or author-prior ansatzes. The derivation chain is therefore self-contained and non-circular.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 1 invented entities

Only the abstract is available, so the ledger is populated at the highest level from the stated proposal. The central claim rests on the existence and effectiveness of an unspecified compact vector representation.

invented entities (1)

compact fixed-size vectorized representation no independent evidence
purpose: To encode textual and visual information of entities and mentions for fast and storage-efficient matching
Introduced in the abstract as the core novel component enabling the efficiency claims; no independent evidence or construction details provided.

pith-pipeline@v0.9.1-grok · 5682 in / 1160 out tokens · 24848 ms · 2026-06-27T08:34:33.789094+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

23 extracted references · 14 canonical work pages · 1 internal anchor

[1]

Zhiwei Hu, Víctor Gutiérrez-Basulto, Ru Li, and Jeff Z. Pan. 2025. Multi-level Matching Network for Multimodal Entity Linking. InProceedings of the 31st ACM SIGKDD conference on Knowledge Discovery and Data Mining (KDD’25). Association for Computing Machinery, New York, NY, 508–519. doi:10.1145/ 3690624.3709306

arXiv 2025
[2]

Zhiwei Hu, Víctor Gutiérrez-Basulto, Zhiliang Xiang, Ru Li, and Jeff Z. Pan. 2025. Multi-level Mixture of Experts for Multimodal Entity Linking. InProceedings of the 31st ACM SIGKDD conference on Knowledge Discovery and Data Mining (KDD’25), Vol. 2. Association for Computing Machinery, New York, NY, 979–990. doi:10.1145/3711896.3737060

work page doi:10.1145/3711896.3737060 2025
[3]

Juyeon Kim, Geon Lee, Taeuk Kim, and Kijung Shin. 2025. KGMEL: Knowledge Graph-enhanced Multimodal Entity Linking. InProceedings of the 48th interna- tional ACM SIGIR conference on Research and Development in Information Retrieval (SIGIR 2025). Association for Computing Machinery, New York, NY, 3015–3019. doi:10.1145/3726302.3730217

work page doi:10.1145/3726302.3730217 2025
[4]

Qi Liu, Yongyi He, Tong Xu, Defu Lian, Che Liu, Zhi Zheng, and Enhong Chen
[5]

Hique: Hierarchical question embedding network for multimodal depression detection,

UniMEL: A Unified Framework for Multimodal Entity Linking with Large Language Models. InProceedings of the 33rd ACM international Conference on Information and Knowledge Management (CIKM’24). Association for Computing Machinery, New York, NY, 1909–1919. doi:10.1145/3627673.3679793

work page doi:10.1145/3627673.3679793 1909
[6]

Ziyan Liu, Junwen Li, Kaiwen Li, Tong Ruan, Chao Wang, Xinyan He, Zongyu Wang, Xuezhi Cao, and Jingping Liu. 2025. I2CR: Intra- and Inter-modal Col- laborative Reflections for Multimodal Entity Linking. InProceedings of the 33rd ACM international conference on Multimedia (MM’25). Association for Computing Machinery, New York, NY, 4942–4951. doi:10.1145/37...

work page doi:10.1145/3746027.3755674 2025
[7]

Shayne Longpre, Kartik Perisetla, Anthony Chen, Nikhil Ramesh, Chris DuBois, and Sameer Singh. 2021. Entity-based Knowledge Conflicts in Question An- swering. InProceedings of the 2021 conference on Empirical Methods in Natural Language Processing (EMNLP 2021). Association for Computational Linguistics, 7052–7063. doi:10.18653/v1/2021.emnlp-main.565

work page doi:10.18653/v1/2021.emnlp-main.565 2021
[8]

Pengfei Luo, Tong Xu, Che Liu, Suojuan Zhang, Linli Xu, Minglei Li, and Enhong Chen. 2024. Bridging Gaps in Content and Knowledge for Multimodal Entity Linking. InProceedings of the 32nd ACM international conference on Multimedia (MM’24). Association for Computing Machinery, New York, NY, 9311–9320. doi:10. 1145/3664647.3681661

arXiv 2024
[9]

Pengfei Luo, Tong Xu, Shiwei Wu, Chen Zhu, Linli Xu, and Enhong Chen. 2023. Multi-grained Multimodal Interaction Network for Entity Linking. InProceedings of the 29th ACM SIGKDD conference on Knowledge Discovery and Data Mining (KDD’23). Association for Computing Machinery, New York, NY, 1583–1594. doi:10.1145/3580305.3599439

work page doi:10.1145/3580305.3599439 2023
[10]

Edgar Meij, Krisztian Balog, and Daan Odijk. 2014. Entity Linking and Retrieval for Semantic Search. InProceedings of the 7th ACM international conference on Web Search and Data Mining (WSDM’14). Association for Computing Machinery, New York, NY, 683–684. doi:10.1145/2556195.2556201

work page doi:10.1145/2556195.2556201 2014
[11]

Seungwhan Moon, Leonardo Neves, and Vitor Carvalho. 2018. Multimodal Named Entity Disambiguation for Noisy Social Media Posts. InProceedings of the 56th annual meeting of the Association for Computational Linguistics (ACL 2018). Association for Computational Linguistics, 2000–2008. doi:10.18653/v1/P18-1186

work page doi:10.18653/v1/p18-1186 2018
[12]

Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. 2021. Learning Transferable Visual Models from Natural Language Supervision. InProceedings of the 38th International Conference on Machine Learning (ICML 2021), Vol. 139. ...

2021
[13]

Nils Reimers and Iryna Gurevych. 2019. Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks. InProceedings of the 2019 conference on Em- pirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP 2019). Association for Computational Linguistics, 3982–3992. doi:10.18653/v...

work page doi:10.18653/v1/d19-1410 2019
[14]

Robertson, Steve Walker, Susan Jones, Micheline Hancock-Beaulieu, and Mike Gatford

Stephen E. Robertson, Steve Walker, Susan Jones, Micheline Hancock-Beaulieu, and Mike Gatford. 1994. Okapi at TREC-3. InProceedings of the 3rd Text Retrieval Conference (TREC-3). Department of Commerce, NIST, 109–126. https://trec.nist. gov/pubs/trec3/papers/city.ps.gz

1994
[15]

Senbao Shi, Zhenran Xu, Baotian Hu, and Min Zhang. 2024. Generative Multi- modal Entity Linking. InProceedings of the 2024 joint international conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024). ELRA and ICCL, 7654–7665. https://aclanthology.org/2024.lrec-main.676/

2024
[16]

Xuhui Sui, Ying Zhang, Yu Zhao, Kehui Song, Baohang Zhou, and Xiaojie Yuan
[17]

InProceedings of Findings of the Association for Computational Linguistics (ACL 2024)

MELOV: Multimodal Entity Linking with Optimized Visual Features in Latent Space. InProceedings of Findings of the Association for Computational Linguistics (ACL 2024). Association for Computational Linguistics, 816–826. doi:10. 18653/v1/2024.findings-acl.46

2024
[18]

Representation Learning with Contrastive Predictive Coding

Aäron van den Oord, Yazhe Li, and Oriol Vinyals. 2018.Representation Learning with Contrastive Predictive Coding. arXiv:1807.03748 doi:10.48550/arXiv.1807. 03748

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1807 2018
[19]

Peng Wang, Jiangheng Wu, and Xiaohang Chen. 2022. Multimodal Entity Link- ing with Gated Hierarchical Fusion and Contrastive Training. InProceedings of the 45th international ACM SIGIR conference on Research and Development in Information Retrieval (SIGIR 2022). Association for Computing Machinery, New York, NY, 938–948. doi:10.1145/3477495.3531867

work page doi:10.1145/3477495.3531867 2022
[20]

Xuwu Wang, Junfeng Tian, Min Gui, Zhixu Li, Rui Wang, Ming Yan, Lihan Chen, and Yanghua Xiao. 2022. WikiDiverse: A Multimodal Entity Linking Dataset with Diversified Contextual Topics and Entity Types. InProceedings of the 60th annual meeting of the Association for Computational Linguistics (ACL 2022). Association for Computational Linguistics, 4785–4797....

work page doi:10.18653/v1/2022.acl-long.328 2022
[21]

Wenhan Xiong, Mo Yu, Shiyu Chang, Xiaoxiao Guo, and William Yang Wang
[22]

InProceedings of the 57th annual meeting of the Association for Computational Linguistics (ACL 2019)

Improving Question Answering over Incomplete KBs with Knowledge- Aware Reader. InProceedings of the 57th annual meeting of the Association for Computational Linguistics (ACL 2019). Association for Computational Linguistics, 4258–4264. doi:10.18653/v1/P19-1417

work page doi:10.18653/v1/p19-1417 2019
[23]

Zefeng Zhang, Jiawei Sheng, Chuang Zhang, Liangyunzhi Liangyunzhi, Wenyuan Zhang, Siqi Wang, and Tingwen Liu. 2024. Optimal Transport Guided Correlation Assignment for Multimodal Entity Linking. InProceedings of Findings of the Asso- ciation for Computational Linguistics (ACL 2024). Association for Computational Linguistics, 4103–4117. doi:10.18653/v1/202...

work page doi:10.18653/v1/2024.findings-acl.243 2024

[1] [1]

Zhiwei Hu, Víctor Gutiérrez-Basulto, Ru Li, and Jeff Z. Pan. 2025. Multi-level Matching Network for Multimodal Entity Linking. InProceedings of the 31st ACM SIGKDD conference on Knowledge Discovery and Data Mining (KDD’25). Association for Computing Machinery, New York, NY, 508–519. doi:10.1145/ 3690624.3709306

arXiv 2025

[2] [2]

Zhiwei Hu, Víctor Gutiérrez-Basulto, Zhiliang Xiang, Ru Li, and Jeff Z. Pan. 2025. Multi-level Mixture of Experts for Multimodal Entity Linking. InProceedings of the 31st ACM SIGKDD conference on Knowledge Discovery and Data Mining (KDD’25), Vol. 2. Association for Computing Machinery, New York, NY, 979–990. doi:10.1145/3711896.3737060

work page doi:10.1145/3711896.3737060 2025

[3] [3]

Juyeon Kim, Geon Lee, Taeuk Kim, and Kijung Shin. 2025. KGMEL: Knowledge Graph-enhanced Multimodal Entity Linking. InProceedings of the 48th interna- tional ACM SIGIR conference on Research and Development in Information Retrieval (SIGIR 2025). Association for Computing Machinery, New York, NY, 3015–3019. doi:10.1145/3726302.3730217

work page doi:10.1145/3726302.3730217 2025

[4] [4]

Qi Liu, Yongyi He, Tong Xu, Defu Lian, Che Liu, Zhi Zheng, and Enhong Chen

[5] [5]

Hique: Hierarchical question embedding network for multimodal depression detection,

UniMEL: A Unified Framework for Multimodal Entity Linking with Large Language Models. InProceedings of the 33rd ACM international Conference on Information and Knowledge Management (CIKM’24). Association for Computing Machinery, New York, NY, 1909–1919. doi:10.1145/3627673.3679793

work page doi:10.1145/3627673.3679793 1909

[6] [6]

Ziyan Liu, Junwen Li, Kaiwen Li, Tong Ruan, Chao Wang, Xinyan He, Zongyu Wang, Xuezhi Cao, and Jingping Liu. 2025. I2CR: Intra- and Inter-modal Col- laborative Reflections for Multimodal Entity Linking. InProceedings of the 33rd ACM international conference on Multimedia (MM’25). Association for Computing Machinery, New York, NY, 4942–4951. doi:10.1145/37...

work page doi:10.1145/3746027.3755674 2025

[7] [7]

Shayne Longpre, Kartik Perisetla, Anthony Chen, Nikhil Ramesh, Chris DuBois, and Sameer Singh. 2021. Entity-based Knowledge Conflicts in Question An- swering. InProceedings of the 2021 conference on Empirical Methods in Natural Language Processing (EMNLP 2021). Association for Computational Linguistics, 7052–7063. doi:10.18653/v1/2021.emnlp-main.565

work page doi:10.18653/v1/2021.emnlp-main.565 2021

[8] [8]

Pengfei Luo, Tong Xu, Che Liu, Suojuan Zhang, Linli Xu, Minglei Li, and Enhong Chen. 2024. Bridging Gaps in Content and Knowledge for Multimodal Entity Linking. InProceedings of the 32nd ACM international conference on Multimedia (MM’24). Association for Computing Machinery, New York, NY, 9311–9320. doi:10. 1145/3664647.3681661

arXiv 2024

[9] [9]

Pengfei Luo, Tong Xu, Shiwei Wu, Chen Zhu, Linli Xu, and Enhong Chen. 2023. Multi-grained Multimodal Interaction Network for Entity Linking. InProceedings of the 29th ACM SIGKDD conference on Knowledge Discovery and Data Mining (KDD’23). Association for Computing Machinery, New York, NY, 1583–1594. doi:10.1145/3580305.3599439

work page doi:10.1145/3580305.3599439 2023

[10] [10]

Edgar Meij, Krisztian Balog, and Daan Odijk. 2014. Entity Linking and Retrieval for Semantic Search. InProceedings of the 7th ACM international conference on Web Search and Data Mining (WSDM’14). Association for Computing Machinery, New York, NY, 683–684. doi:10.1145/2556195.2556201

work page doi:10.1145/2556195.2556201 2014

[11] [11]

Seungwhan Moon, Leonardo Neves, and Vitor Carvalho. 2018. Multimodal Named Entity Disambiguation for Noisy Social Media Posts. InProceedings of the 56th annual meeting of the Association for Computational Linguistics (ACL 2018). Association for Computational Linguistics, 2000–2008. doi:10.18653/v1/P18-1186

work page doi:10.18653/v1/p18-1186 2018

[12] [12]

Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. 2021. Learning Transferable Visual Models from Natural Language Supervision. InProceedings of the 38th International Conference on Machine Learning (ICML 2021), Vol. 139. ...

2021

[13] [13]

Nils Reimers and Iryna Gurevych. 2019. Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks. InProceedings of the 2019 conference on Em- pirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP 2019). Association for Computational Linguistics, 3982–3992. doi:10.18653/v...

work page doi:10.18653/v1/d19-1410 2019

[14] [14]

Robertson, Steve Walker, Susan Jones, Micheline Hancock-Beaulieu, and Mike Gatford

Stephen E. Robertson, Steve Walker, Susan Jones, Micheline Hancock-Beaulieu, and Mike Gatford. 1994. Okapi at TREC-3. InProceedings of the 3rd Text Retrieval Conference (TREC-3). Department of Commerce, NIST, 109–126. https://trec.nist. gov/pubs/trec3/papers/city.ps.gz

1994

[15] [15]

Senbao Shi, Zhenran Xu, Baotian Hu, and Min Zhang. 2024. Generative Multi- modal Entity Linking. InProceedings of the 2024 joint international conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024). ELRA and ICCL, 7654–7665. https://aclanthology.org/2024.lrec-main.676/

2024

[16] [16]

Xuhui Sui, Ying Zhang, Yu Zhao, Kehui Song, Baohang Zhou, and Xiaojie Yuan

[17] [17]

InProceedings of Findings of the Association for Computational Linguistics (ACL 2024)

MELOV: Multimodal Entity Linking with Optimized Visual Features in Latent Space. InProceedings of Findings of the Association for Computational Linguistics (ACL 2024). Association for Computational Linguistics, 816–826. doi:10. 18653/v1/2024.findings-acl.46

2024

[18] [18]

Representation Learning with Contrastive Predictive Coding

Aäron van den Oord, Yazhe Li, and Oriol Vinyals. 2018.Representation Learning with Contrastive Predictive Coding. arXiv:1807.03748 doi:10.48550/arXiv.1807. 03748

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1807 2018

[19] [19]

Peng Wang, Jiangheng Wu, and Xiaohang Chen. 2022. Multimodal Entity Link- ing with Gated Hierarchical Fusion and Contrastive Training. InProceedings of the 45th international ACM SIGIR conference on Research and Development in Information Retrieval (SIGIR 2022). Association for Computing Machinery, New York, NY, 938–948. doi:10.1145/3477495.3531867

work page doi:10.1145/3477495.3531867 2022

[20] [20]

Xuwu Wang, Junfeng Tian, Min Gui, Zhixu Li, Rui Wang, Ming Yan, Lihan Chen, and Yanghua Xiao. 2022. WikiDiverse: A Multimodal Entity Linking Dataset with Diversified Contextual Topics and Entity Types. InProceedings of the 60th annual meeting of the Association for Computational Linguistics (ACL 2022). Association for Computational Linguistics, 4785–4797....

work page doi:10.18653/v1/2022.acl-long.328 2022

[21] [21]

Wenhan Xiong, Mo Yu, Shiyu Chang, Xiaoxiao Guo, and William Yang Wang

[22] [22]

InProceedings of the 57th annual meeting of the Association for Computational Linguistics (ACL 2019)

Improving Question Answering over Incomplete KBs with Knowledge- Aware Reader. InProceedings of the 57th annual meeting of the Association for Computational Linguistics (ACL 2019). Association for Computational Linguistics, 4258–4264. doi:10.18653/v1/P19-1417

work page doi:10.18653/v1/p19-1417 2019

[23] [23]

Zefeng Zhang, Jiawei Sheng, Chuang Zhang, Liangyunzhi Liangyunzhi, Wenyuan Zhang, Siqi Wang, and Tingwen Liu. 2024. Optimal Transport Guided Correlation Assignment for Multimodal Entity Linking. InProceedings of Findings of the Asso- ciation for Computational Linguistics (ACL 2024). Association for Computational Linguistics, 4103–4117. doi:10.18653/v1/202...

work page doi:10.18653/v1/2024.findings-acl.243 2024