Pith. sign in

REVIEW 4 major objections 3 minor 42 references

On Evaluating the Adversarial Robustness of Foundation Models for Multimodal Entity Linking

T0 review · 4 major / 3 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Multimodal entity linking models are broadly vulnerable to visual adversarial perturbations, and textual context only partially rescues them; retrieval-augmented LLM-RetLink improves accuracy under attack by 0.4–35.7%.

desk verdict Potentially the first MEL adversarial robustness study, but the full text is unreadable and the retrieval-leakage risk is unresolved; referee only after a clean, filtered-report. read the letter →

arxiv 2508.15481 v1 pith:T2BJAK6O submitted 2025-08-21 cs.IR

Fang Wang , Yongjie Wang , Zonghao Yang , Minghao Hu , Xiaoying Bai This is my paper
classification cs.IR
keywords multimodalentitylinkingadversarialrobustnessvisualperturbationretrieval-augmentedgenerationlargevisionmodelsImage-to-TextImage+Text-to-Textdataset
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper asks whether mainstream multimodal entity linking (MEL) systems can survive small, targeted visual perturbations. It reports that they largely cannot: across two core MEL tasks, Image-to-Text and Image+Text-to-Text, current models lose accuracy when the image is adversarially modified. The paper also finds that surrounding textual context partially offsets the damage, and uses that insight to build LLM-RetLink, a two-stage system that first generates entity descriptions with a large vision model and then pulls candidate descriptive sentences from the web. On five datasets, LLM-RetLink improves MEL accuracy by 0.4% to 35.7%, with the largest gains in adversarial settings. The paper also releases the first MEL adversarial example dataset, giving future work a concrete benchmark for robustness.

What carries the argument

The load-bearing mechanism is LLM-RetLink's two-stage architecture: stage one uses a large vision model (LVM) to generate an initial description of the visual entity, and stage two retrieves candidate descriptive sentences from the web to provide complementary textual context. That context is what the paper identifies as the partial antidote to visual perturbation. The evaluation itself rests on a newly constructed MEL adversarial example dataset that perturbs images while keeping the text input intact, which is what makes the measured accuracy drops attributable to the visual channel.

What would settle it

Run LLM-RetLink on the released adversarial dataset with the web-retrieval stage ablated, replacing retrieved sentences with random or non-informative text; if accuracy under attack drops back to baseline levels, the reported robustness gain is attributable to retrieval access to the answer rather than to anti-interference ability. Also, measure whether retrieved candidate sentences contain the ground-truth entity's name or identifying phrase for attacked examples.

Watch

Extended reading notes

Core claim

The paper claims that multimodal entity linking models lack sufficient robustness against visual perturbations, and that this is a previously unexamined weakness of the MEL pipeline. It constructs the first MEL adversarial example dataset and runs the first systematic robustness evaluation across two tasks, I2T and IT2T. The central positive result is that contextual semantic information in the text input can partially absorb the effect of visual attacks. Building on that, the proposed LLM-RetLink uses large vision models to extract initial entity descriptions and web retrieval to dynamically generate candidate descriptive sentences; the paper reports accuracy improvements of 0.4% to 35.7% o

Load-bearing premise

The load-bearing assumption is that the web-retrieval stage of LLM-RetLink supplies candidate descriptive sentences without exposing the ground-truth entity, and that the visual perturbation attacks used in the evaluation are a faithful proxy for adversarial robustness in real MEL use.

Editorial extensions

If this is right

  • If the central claim holds, MEL deployments that rely on visual input alone will be fragile against small adversarial image changes, so robustness needs to be a design criterion, not a post-hoc patch.
  • Because textual context partially restores accuracy, adding a retrieval or context-expansion stage is a concrete defense direction for multimodal linking and adjacent tasks.
  • The released MEL adversarial dataset gives subsequent work a shared testbed, so robustness results across different MEL models can be compared directly.
  • The 0.4% to 35.7% accuracy improvements reported for LLM-RetLink suggest that the gain under attack can be much larger than the gain in clean settings, making the method a specifically adversarial defense.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial extension: the paper's gains would be overstated if the web-retrieval stage leaks the ground-truth entity; a clean test is to run LLM-RetLink with retrieval disabled or with retrieved sentences replaced by irrelevant text and measure how much of the improvement survives.
  • Editorial extension: the same context-suppresses-perturbation mechanism may transfer to other vision-language tasks such as visual question answering or visual grounding, where small image changes are also known to flip predictions.
  • Editorial extension: adaptive attacks that are designed with LLM-RetLink's retrieval stage in mind may erode the reported margins; evaluating against an adversary that can also manipulate retrieved text would be a stricter test.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 3 minor

Summary. The paper claims to present the first comprehensive evaluation of visual adversarial robustness for multimodal entity linking (MEL), covering Image-to-Text and Image+Text-to-Text tasks. The abstract states that mainstream MEL models lack robustness against visual perturbations, that contextual semantic information can partially mitigate such perturbations, and that the proposed LLM-RetLink—which combines large vision model (LVM) description extraction with web-based retrieval of candidate descriptive sentences—improves MEL accuracy by 0.4% to 35.7%, with larger gains under adversarial conditions. The abstract also announces the release of a new MEL adversarial example dataset. However, the body of the submitted manuscript is an unreadable, encoding-corrupted stream, and the only identifiable text after the abstract is a line referencing a different arXiv identifier and subject classification.

Significance. If the empirical claims are correct, the paper would provide a useful benchmark (the first MEL adversarial dataset) and an actionable finding about the mitigating role of context in adversarial settings. The proposed retrieval-augmented method could be of practical interest, provided the reported gains are not an artifact of test-time leakage. The significance is conditional on the methods and experiments being available for inspection; at present, the central claims are unverifiable from the submission alone.

major comments (4)
  1. [Full text (after Abstract)] The body of the manuscript is an unreadable encoding-corrupted stream; no methods, experimental setup, tables, or per-dataset results can be inspected. The central empirical claims—about model vulnerability, context mitigation, and the LLM-RetLink improvements—are supported only by the abstract. Please provide a readable manuscript with full methods, dataset descriptions, attack configurations, baseline details, and results tables, and cite the specific sections for each claim.
  2. [Abstract, LLM-RetLink method description] The abstract states that LLM-RetLink uses 'web-based retrieval' to generate candidate descriptive sentences. In standard MEL benchmarks, the gold entity is a named Wikipedia entity. If retrieved web snippets are not filtered to exclude the gold entity, the retrieval stage can leak the ground-truth name, especially under adversarial perturbation when the visual signal is degraded. This would inflate the reported 0.4%–35.7% improvements. No leakage-prevention mechanism is mentioned in the abstract. Please specify the retrieval query construction, any filtering of retrieved content against the gold entity, and provide an ablation restricted to cases where the retrieved text does not contain the gold entity's name.
  3. [Abstract, results claims] The reported improvement range of 0.4%–35.7% is extremely wide, and the abstract provides no per-dataset breakdown, error bars, or significance tests. The phrase 'significantly improves' is therefore unsupported. Please present per-dataset results with variance measures, statistical tests where appropriate, and a discussion of why the gains vary so widely across datasets and conditions.
  4. [Last line of full text] The final line of the manuscript reads 'arXiv:2508.15482v2 [hep-ph] 6 Feb 2026', which does not match the stated identifier 'arXiv:2508.15481 (cs.IR)'. This is inconsistent and suggests that the uploaded file is corrupted, incorrect, or contains extraneous content. Please verify that the submitted file is the correct manuscript and that all text is properly rendered.
minor comments (3)
  1. [Abstract] The abstract claims this is the 'first comprehensive evaluation' of MEL robustness; the full text should provide a thorough related-work comparison that substantiates this claim.
  2. [Abstract] The term 'LVM' is introduced without expansion in the abstract; define it as 'large vision model' on first use for readability.
  3. [Abstract] The word 'significantly' is used in a non-statistical sense; clarify the intended meaning to avoid confusion with hypothesis testing.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation identified; the claims are empirical benchmark results and a proposed method, not logical reductions to their own inputs.

full rationale

The paper, as available, consists of an abstract and an unreadable encoding-corrupted body. The abstract reports (1) an empirical evaluation showing that MEL models are vulnerable to visual perturbations, (2) an observation that contextual semantic information partially mitigates the impact, and (3) a proposed method LLM-RetLink that improves accuracy by 0.4%-35.7%. None of these claims is framed as a derivation from a premise that already contains the conclusion. There is no quoted equation, fitted parameter renamed as a prediction, uniqueness theorem imported from the authors' prior work, or ansatz smuggled in via citation that I can exhibit. The main correctness risk noted by the skeptic is test-time leakage from web-based retrieval: if retrieved descriptive sentences contain the gold entity name, the reported gains may reflect answer exposure rather than visual anti-interference. That is a serious experimental-validity concern, but it is not circularity under the stated hard rules: it is not a self-definitional reduction, not a fitted-input-called-prediction, and not a self-citation load-bearing argument. Because no specific circular step can be quoted and exhibited, the appropriate finding is no significant circularity, score 0.

Assumptions & free parameters 1 free parameters · 2 assumptions · 0 invented entities

Abstract-only review. No new physical or conceptual entities are introduced; the proposed LLM-RetLink is a method, and the MEL adversarial dataset is a benchmark artifact rather than an invented entity. Free parameters and assumptions listed are the ones inferable from the abstract; the unreadable full text prevents an exhaustive audit.

free parameters (1)
  • LLM-RetLink hyperparameters (retrieval top-k, LVM prompt settings, LLM inference settings)
    Not specified in the abstract. The reported gains range from 0.4% to 35.7% across five datasets, so per-dataset configuration is likely material to the result and is unverifiable here.
assumptions (2)
  • domain assumption Standard visual adversarial attack algorithms and perturbation budgets transfer to multimodal entity linking models.
    The abstract assumes that adversarial robustness of MEL can be measured by applying visual perturbations; the specific attacks, budgets, and their transferability are not readable.
  • domain assumption Web-based retrieval returns descriptively useful candidate sentences without exposing ground-truth entity labels.
    The claimed gains of LLM-RetLink depend on retrieval quality and on non-leakage of the answer; neither can be verified from the abstract.

how reviews work

0 comments
Cite this review

Pith. "Pith review of On Evaluating the Adversarial Robustness of Foundation Models for Multimodal Entity Linking." pith.science (2026). https://pith.science/paper/T2BJAK6O

@misc{pith2026250815481,
  author       = {Pith},
  title        = {Pith review of: On Evaluating the Adversarial Robustness of Foundation Models for Multimodal Entity Linking},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/T2BJAK6O}},
  note         = {Machine review of arXiv:2508.15481}
}
read the original abstract

The explosive growth of multimodal data has driven the rapid development of multimodal entity linking (MEL) models. However, existing studies have not systematically investigated the impact of visual adversarial attacks on MEL models. We conduct the first comprehensive evaluation of the robustness of mainstream MEL models under different adversarial attack scenarios, covering two core tasks: Image-to-Text (I2T) and Image+Text-to-Text (IT2T). Experimental results show that current MEL models generally lack sufficient robustness against visual perturbations. Interestingly, contextual semantic information in input can partially mitigate the impact of adversarial perturbations. Based on this insight, we propose an LLM and Retrieval-Augmented Entity Linking (LLM-RetLink), which significantly improves the model's anti-interference ability through a two-stage process: first, extracting initial entity descriptions using large vision models (LVMs), and then dynamically generating candidate descriptive sentences via web-based retrieval. Experiments on five datasets demonstrate that LLM-RetLink improves the accuracy of MEL by 0.4%-35.7%, especially showing significant advantages under adversarial conditions. This research highlights a previously unexplored facet of MEL robustness, constructs and releases the first MEL adversarial example dataset, and sets the stage for future work aimed at strengthening the resilience of multimodal systems in adversarial environments.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

42 extracted references · 33 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Adjali, O.; Besan c on, R.; Ferret, O.; Le Borgne, H.; and Grau, B. 2020. Building a Multimodal Entity Linking Dataset From Tweets. In Proceedings of the 12th Language Resources and Evaluation Conference, 4285--4292. Marseille, France: European Language Resources Association. ISBN 979-10-95546-34-4

    work page 2020

  4. [4]

    Archana, R.; and Jeevaraj, P. E. 2024. Deep learning models for digital image processing: a review. Artificial Intelligence Review, 57(1): 11

    work page 2024

  5. [5]

    Athalye, A.; Carlini, N.; and Wagner, D. 2018. Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examples. In Proceedings of the 35th International Conference on Machine Learning, ICML 2018

    work page 2018

  6. [6]

    Bai, S.; Chen, K.; Liu, X.; Wang, J.; Ge, W.; Song, S.; Dang, K.; Wang, P.; Wang, S.; Tang, J.; Zhong, H.; Zhu, Y.; Yang, M.; Li, Z.; Wan, J.; Wang, P.; Ding, W.; Fu, Z.; Xu, Y.; Ye, J.; Zhang, X.; Xie, T.; Cheng, Z.; Zhang, H.; Yang, Z.; Xu, H.; and Lin, J. 2025. Qwen2.5-VL Technical Report. CoRR, abs/2502.13923

    arXiv 2025

  7. [7]

    Carlini, N.; and Wagner, D. 2017. Towards Evaluating the Robustness of Neural Networks. In 2017 IEEE Symposium on Security and Privacy (SP), 39--57

    work page 2017

  8. [8]

    Chi, L.; Msahli, M.; Zhang, Q.; Qiu, H.; Zhang, T.; Memmi, G.; and Qiu, M. 2024. Adversarial attacks on autonomous driving systems in the physical world: a survey. IEEE Transactions on Intelligent Vehicles

    work page 2024

Show all 42 references
  1. [9]

    Croce, F.; and Hein, M. 2020. Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attacks. In III, H. D.; and Singh, A., eds., Proceedings of the 37th International Conference on Machine Learning, volume 119 of Proceedings of Machine Learni...

    2020

  2. [10]

    K.; and Lim, S.-N

    Cui, X.; Aparcedo, A.; Jang, Y. K.; and Lim, S.-N. 2024. On the robustness of large multimodal models against image adversarial attacks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 24625--24634

    2024

  3. [11]

    Gan, J.; Luo, J.; Wang, H.; Wang, S.; He, W.; and Huang, Q. 2021. Multimodal Entity Linking: a New Dataset and a Baseline. Multimedia

    2021

  4. [12]

    Hu, M.; Dong, H.; Luo, P.; Han, S.; and Zhang, D. 2024. KET-QA: A dataset for knowledge enhanced table question answering. arXiv preprint arXiv:2405.08099

    2024 arXiv

  5. [13]

    Huang, H.; Nie, Z.; Wang, Z.; and Shang, Z. 2024. Cross-modal and uni-modal soft-label alignment for image-text retrieval. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, 18298--18306

    2024

  6. [14]

    Jia, C.; Yang, Y.; Xia, Y.; Chen, Y.-T.; Parekh, Z.; Pham, H.; Le, Q.; Sung, Y.-H.; Li, Z.; and Duerig, T. 2021. Scaling up visual and vision-language representation learning with noisy text supervision. In International conference on machine learning, 4904--4916. PMLR

    2021

  7. [15]

    Li, J.; Li, D.; Savarese, S.; and Hoi, S. 2023. Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models. In International conference on machine learning, 19730--19742. PMLR

    2023

  8. [16]

    Liang, X.; Yang, E.; Deng, C.; and Yang, Y. 2024. CrossFormer: Cross-modal Representation Learning via Heterogeneous Graph Transformer. ACM Transactions on Multimedia Computing, Communications and Applications, 20(12): 1--21

    2024

  9. [17]

    Liu, H.; Li, C.; Wu, Q.; and Lee, Y. J. 2023. Visual Instruction Tuning. In Oh, A.; Naumann, T.; Globerson, A.; Saenko, K.; Hardt, M.; and Levine, S., eds., Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, N...

    2023

  10. [18]

    Long, X.; Zeng, J.; Meng, F.; Zhou, J.; and Zhou, B. 2024. Trust in Internal or External Knowledge? Generative Multi-Modal Entity Linking with Knowledge Retriever. In Ku, L.; Martins, A.; and Srikumar, V., eds., Findings of the Association for Computational Linguistics, ACL 20...

    2024

  11. [19]

    Luo, P.; Xu, T.; Wu, S.; Zhu, C.; Xu, L.; and Chen, E. 2023. Multi-Grained Multimodal Interaction Network for Entity Linking. In Singh, A. K.; Sun, Y.; Akoglu, L.; Gunopulos, D.; Yan, X.; Kumar, R.; Ozcan, F.; and Ye, J., eds., Proceedings of the 29th ACM SIGKDD Conference on ...

    2023

  12. [20]

    Madry, A.; Makelov, A.; Schmidt, L.; Tsipras, D.; and Vladu, A. 2018. Towards Deep Learning Models Resistant to Adversarial Attacks. In International Conference on Learning Representations

    2018

  13. [21]

    Minderer, M.; Gritsenko, A.; Stone, A.; Neumann, M.; Weissenborn, D.; Dosovitskiy, A.; Mahendran, A.; Arnab, A.; Dehghani, M.; Shen, Z.; Wang, X.; Zhai, X.; Kipf, T.; and Houlsby, N. 2022. Simple Open-Vocabulary Object Detection with Vision Transformers. arXiv:2205.06230

    2022 arXiv

  14. [22]

    Moon, S.; Neves, L.; and Carvalho, V. 2018. Multimodal Named Entity Disambiguation for Noisy Social Media Posts. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2000--2008. Melbourne, Australia: Association fo...

    2018

  15. [23]

    B.; and Swami, A

    Papernot, N.; McDaniel, P.; Goodfellow, I.; Jha, S.; Celik, Z. B.; and Swami, A. 2017. Practical Black-Box Attacks against Machine Learning. In Proceedings of the 2017 ACM on Asia Conference on Computer and Communications Security, ASIA CCS '17, 506–519. New York, NY, USA: Ass...

    2017

  16. [24]

    Pons, G.; Bilalli, B.; and Queralt, A. 2024. Knowledge graphs for enhancing large language models in entity disambiguation. In International Semantic Web Conference, 162--179. Springer

    2024

  17. [25]

    T.; Gan, Y.; Saberi, A.; Ozcan, F.; and Arik, S

    Pourreza, M.; Li, H.; Sun, R.; Chung, Y.; Talaei, S.; Kakkar, G. T.; Gan, Y.; Saberi, A.; Ozcan, F.; and Arik, S. O. 2024. Chase-sql: Multi-path reasoning and preference optimized candidate selection in text-to-sql. arXiv preprint arXiv:2410.01943

    2024 arXiv

  18. [26]

    Qiu, J.; Han, W.; Wang, W.; Yang, Z.; Li, L.; Wang, J.; Faloutsos, C.; Li, L.; and Wang, L. 2024. Entity6K: A Large Open-Domain Evaluation Dataset for Real-World Entity Recognition. arXiv preprint arXiv:2403.12339

    2024 arXiv

  19. [27]

    W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; et al

    Radford, A.; Kim, J. W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; et al. 2021. Learning transferable visual models from natural language supervision. In International conference on machine learning, 8748--8763. PMLR

    2021

  20. [28]

    Sima, C.; Renz, K.; Chitta, K.; Chen, L.; Zhang, H.; Xie, C.; Bei wenger, J.; Luo, P.; Geiger, A.; and Li, H. 2024. Drivelm: Driving with graph visual question answering. In European conference on computer vision, 256--274. Springer

    2024

  21. [29]

    Singh, A.; Hu, R.; Goswami, V.; Couairon, G.; Galuba, W.; Rohrbach, M.; and Kiela, D. 2022. Flava: A foundational language and vision alignment model. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 15638--15650

    2022

  22. [30]

    Song, S.; Zhao, S.; Wang, C.; Yan, T.; Li, S.; Mao, X.; and Wang, M. 2024. A Dual-Way Enhanced Framework from Text Matching Point of View for Multimodal Entity Linking. In Wooldridge, M. J.; Dy, J. G.; and Natarajan, S., eds., Thirty-Eighth AAAI Conference on Artificial Intell...

    2024

  23. [31]

    V.; and Sakurai, K

    Su, J.; Vargas, D. V.; and Sakurai, K. 2017. One Pixel Attack for Fooling Deep Neural Networks. IEEE Transactions on Evolutionary Computation, 23: 828--841

    2017

  24. [32]

    Sun, W.; Fan, Y.; Guo, J.; Zhang, R.; and Cheng, X. 2022. Visual Named Entity Linking: A New Dataset and A Baseline. In Goldberg, Y.; Kozareva, Z.; and Zhang, Y., eds., Findings of the Association for Computational Linguistics: EMNLP 2022, Abu Dhabi, United Arab Emirates, Dece...

    2022

  25. [33]

    Szegedy, C.; Zaremba, W.; Sutskever, I.; Bruna, J.; Erhan, D.; Goodfellow, I.; and Fergus, R. 2014. Intriguing properties of neural networks. In International Conference on Learning Representations

    2014

  26. [34]

    Wang, F.; Yin, S.; Bai, X.; Hu, M.; Yan, T.; and Liang, Y. 2025. M \^ 3EL: A Multi-task Multi-topic Dataset for Multi-modal Entity Linking. In Walsh, T.; Shah, J.; and Kolter, Z., eds., AAAI-25, Sponsored by the Association for the Advancement of Artificial Intelligence, Febru...

    2025

  27. [35]

    Wang, M.; Wang, H.; Qi, G.; and Zheng, Q. 2020. Richpedia: A Large-Scale, Comprehensive Multi-Modal Knowledge Graph. Big Data Res., 22: 100159

    2020

  28. [36]

    Wang, X.; Tian, J.; Gui, M.; Li, Z.; Wang, R.; Yan, M.; Chen, L.; and Xiao, Y. 2022. WikiDiverse: A Multimodal Entity Linking Dataset with Diversified Contextual Topics and Entity Types. In Muresan, S.; Nakov, P.; and Villavicencio, A., eds., Proceedings of the 60th Annual Mee...

    2022

  29. [37]

    Xing, S.; Zhao, F.; Wu, Z.; Li, C.; Zhang, J.; and Dai, X. 2023. DRIN: Dynamic Relation Interactive Network for Multimodal Entity Linking. In El - Saddik, A.; Mei, T.; Cucchiara, R.; Bertini, M.; Vallejo, D. P. T.; Atrey, P. K.; and Hossain, M. S., eds., Proceedings of the 31s...

    2023

  30. [38]

    Zhai, X.; Mustafa, B.; Kolesnikov, A.; and Beyer, L. 2023. Sigmoid Loss for Language Image Pre-Training. In IEEE/CVF International Conference on Computer Vision, ICCV 2023, Paris, France, October 1-6, 2023 , 11941--11952. IEEE

    2023

  31. [39]

    Zhang, D.; and Huang, L. 2022. Multimodal Knowledge Learning for Named Entity Disambiguation. In Goldberg, Y.; Kozareva, Z.; and Zhang, Y., eds., Findings of the Association for Computational Linguistics: EMNLP 2022, Abu Dhabi, United Arab Emirates, December 7-11, 2022 , 3160-...

    2022

  32. [40]

    Zhao, Y.; Zhang, Y.; Sui, X.; and Cai, X. 2025. ME3A: A Multimodal Entity Entailment framework for multimodal Entity Alignment. Information Processing & Management, 62(1): 103951

    2025

  33. [41]

    Zhou, X.; Wang, P.; Li, G.; Xie, J.; and Wu, J. 2021. Weibo-MEL, Wikidata-MEL and Richpedia-MEL: Multimodal Entity Linking Benchmark Datasets. In Qin, B.; Jin, Z.; Wang, H.; Pan, J. Z.; Liu, Y.; and An, B., eds., Knowledge Graph and Semantic Computing: Knowledge Graph Empowers...

    2021

  34. [42]

    Zhu, D.; Chen, J.; Shen, X.; Li, X.; and Elhoseiny, M. 2024. MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models. In The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024 . OpenReview.net

    2024

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.