---
title: "ArGuard Shared Task — Alam 2026"
description: "ArGuard is a 2026 shared task on detecting harmful Arabic content in two settings that earlier work treated separately: multimodal memes and prompts sent…"
authors: ["Firoj Alam", "Md. Rafiul Biswas", "Mohamed Bayan Kmainasi", "Ali Ezzat Shahroor", "Hamdy Mubarak", "George Mikros", "Abul Hasnat", "Wajdi Zaghouani"]
year: 2026
venue: "arXiv:2609.29349"
source: "https://arxiv.org/abs/2609.29349"
doi: "10.48550/arxiv.2609.29349"
arxiv: "2609.29349"
licence: "arXiv — non-exclusive license to distribute"
licence_url: "http://arxiv.org/licenses/nonexclusive-distrib/1.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/alam-2026-arguard-shared-task-harmful-content/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/alam-2026-arguard-shared-task-harmful-content/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/2/) / ArGuard Shared Task: Harmful Content Detection in Arabic Memes and LLM Prompts

# ArGuard Shared Task: Harmful Content Detection in Arabic Memes and LLM Prompts

**Authors:** [Firoj Alam](https://md.research.elamur.ai/free/en/authors/firoj-alam/), [Md. Rafiul Biswas](https://md.research.elamur.ai/free/en/authors/md-rafiul-biswas/), [Mohamed Bayan Kmainasi](https://md.research.elamur.ai/free/en/authors/mohamed-bayan-kmainasi/), [Ali Ezzat Shahroor](https://md.research.elamur.ai/free/en/authors/ali-ezzat-shahroor/), [Hamdy Mubarak](https://md.research.elamur.ai/free/en/authors/hamdy-mubarak/), [George Mikros](https://md.research.elamur.ai/free/en/authors/george-mikros/), [Abul Hasnat](https://md.research.elamur.ai/free/en/authors/abul-hasnat/), [Wajdi Zaghouani](https://md.research.elamur.ai/free/en/authors/wajdi-zaghouani/) · **Year:** 2026 · **Venue:** arXiv:2609.29349 · **Licence:** [arXiv — non-exclusive license to distribute](http://arxiv.org/licenses/nonexclusive-distrib/1.0/)

**Links:** [Publisher's page](https://arxiv.org/abs/2609.29349) · [PDF](https://arxiv.org/pdf/2609.29349) · [DOI 10.48550/arxiv.2609.29349](https://doi.org/10.48550/arxiv.2609.29349) · [arXiv 2609.29349](https://arxiv.org/abs/2609.29349) · [OpenAlex](https://openalex.org/W7214391319)

**Citations:** 0 citations (openalex-snapshot · 2026-10-14)

## Summary
ArGuard is a 2026 shared task on detecting harmful Arabic content in two settings that earlier work treated separately: multimodal memes and prompts sent to LLMs. Track A uses AHA-Memes, 5,000 manually annotated Arabic memes, each paired with OCR text. Its subtasks are binary hate detection (A1) and multi-label classification over a ten-label taxonomy (A2). That taxonomy covers attack types plus non-hateful phenomena such as Humor and Sarcasm. Track B asks two questions about Arabic prompts in MSA and dialects: is the prompt safe (B1), and which of seven harm categories does it belong to (B2)? The organizers released baselines, format checkers and scorers, then compared the submitted systems. Of 58 registered teams, 35 took part in the final phase. Binary detection proved far easier than fine-grained labeling. Fine-grained memes were the hardest setting, with a best macro-F1 of only 0.419. Rare labels and train–test distribution shift explain much of that gap. Strong systems shared three habits: ensembling, class-aware training and threshold calibration. The work matters because it gives Arabic content-safety research a shared, reproducible benchmark that goes beyond coarse labels.

## Key points
- Participation: 58 teams registered, 35 submitted final runs and 27 wrote system papers. Model families included AraBERT, Jais, Qwen3-VL, VLM ensembles and lexical methods.
- Best macro-F1 per subtask:
  - A1: 0.823
  - A2: 0.419 (micro-F1 0.576)
  - B1: 0.984, shared by AK_QMUL with a fine-tuned Jais decoder and Lattice AI with an Arabic encoder ensemble
  - B2: 0.790, against a baseline of 0.687, with the top two systems only 0.002 apart
- AHA-Memes data:
  - Splits: 3,500 train, 500 dev, 500 dev_test and a 500-meme blind test set; 37.8% of labeled memes are Hateful.
  - Annotator agreement (Cohen's κ) reaches 0.91 on the hate vs. not-hate decision, falls to 0.75 when annotators assign the type of hate, and drops to 0.67 on the non-hateful subtypes.
- Label imbalance in A2 is extreme: Mocking is the most frequent attack type and Exclusion the rarest. For B1, unsafe prompts make up about 80% of train/dev but only 49% of the test set.
- Release terms: AHA-Memes is on Hugging Face (QCRI/AHA-MEMES) for non-commercial research use, while the Track B data requires a controlled-access request.

## Interesting ideas
- Tuning a separate decision threshold for each label, together with decoding that respects the label hierarchy, helped with rare labels and keeps fine labels consistent with the coarse verdict. — For Mova (frames): irony and speech-act labels could get per-label calibrated thresholds, with a constraint that a fine subtype only fires when its parent act is predicted.
- The taxonomy includes explicit non-hateful subtypes (Humor, Sarcasm), so benign look-alikes become named classes instead of a vague negative. — For Mova (frames): irony detection could model "sounds hostile but is a joke" as a positive category with its own evidence, which also makes the decision easier to explain.
- Correcting for class priors after the label distribution shifts between training and test (unsafe 80% vs 49%) recovered robustness that raw representations alone did not provide. — For Mova (learn): when silver data agreed by witness parsers has different label proportions than target texts, apply prior correction and report results under both distributions.
- Macro-F1 that weights every label equally, combined with per-class F1 reporting, exposed failures on rare classes such as Exclusion that accuracy would hide. — For Mova (learn): evaluate semantic roles or frame types with per-class scores so that rare roles cannot vanish behind frequent ones.
- Per-level inter-annotator agreement (0.91 binary, 0.67 for subtypes) gives a realistic ceiling for each granularity before any model is judged. — For Mova (learn): measure agreement between LLM teacher tiers at each label depth and use it as the target ceiling for the student.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

Fields: [Linguistics & languages](https://md.research.elamur.ai/free/en/shelves/linguistics/)

## Cite

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@misc{alam2026arguard,
  title = {{ArGuard Shared Task: Harmful Content Detection in Arabic Memes and LLM Prompts}},
  author = {Alam, Firoj and Biswas, Md. Rafiul and Kmainasi, Mohamed Bayan and Shahroor, Ali Ezzat and Mubarak, Hamdy and Mikros, George and Hasnat, Abul and Zaghouani, Wajdi},
  year = {2026},
  doi = {10.48550/arxiv.2609.29349},
  eprint = {2609.29349},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2609.29349}
}
```

- [← previous ArgAssist: LLM-based Argument Synthesis for Insurance Disputes](https://md.research.elamur.ai/free/en/papers/sinha-2026-argassist-llm-based-argument-synthesis/)
- [next → Armenian AutoEpiDoc: Automated Extraction and Encoding of Armenian Inscriptions into EpiDoc TEI/XML](https://md.research.elamur.ai/free/en/papers/tamrazyan-2026-armenian-autoepidoc-automated-extraction-encoding/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/2/)

## References

57 references, 7 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Qaiser Abbas, Muhammad Irzam, and Jens Schneider. 2026. Qysr at ArGuard Shared Tasks: Arabic harmful prompt detection via sparse residual adapter. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Qaiser+Abbas%2C+Muhammad+Irzam%2C+and+Jens+Schneider%2E+2026%2E+Qysr+at+ArGuard+Shared+Tasks%3A+Arabic+harmful+prompt+detection+via+sparse+residual+adapter%2E+In+Proceedings+of+the+Fourth+Arabic+Natural+Language+)
2. Azza Abouzied, Firoj Alam, Raian Ali, and Paolo Papotti. 2025. Combating misinformation in the arab world: Challenges and opportunities. Communications of the ACM, 68(10):48–53. [find](https://scholar.google.com/scholar?q=Azza+Abouzied%2C+Firoj+Alam%2C+Raian+Ali%2C+and+Paolo+Papotti%2E+2025%2E+Combating+misinformation+in+the+arab+world%3A+Challenges+and+opportunities%2E+Communications+of+the+ACM%2C+68%2810%29%3A48%E2%80%9353%2E)
3. Olaronke Oluwayemisi Adebanji, Emmanuel O. Ilesanmi, Olumide Ebenezer Ojo, Hiram Calvo, Grigori Sidorov, and Anna Feldman. 2026. LingLab at Ar-Guard Shared Tasks: A lightweight hybrid classifier for Arabic harmful prompt detection. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Olaronke+Oluwayemisi+Adebanji%2C+Emmanuel+O%2E+Ilesanmi%2C+Olumide+Ebenezer+Ojo%2C+Hiram+Calvo%2C+Grigori+Sidorov%2C+and+Anna+Feldman%2E+2026%2E+LingLab+at+Ar%2DGuard+Shared+Tasks%3A+A+lightweight+hybrid+classifier+for+A)
4. Mohammed Al-Ani and Abdelrahman Sameh Ibrahim Ibrahim Hassaan. 2026. DetectorAI at ArGuard Shared Tasks: Calibrated multimodal and label-frequency-aware systems for Arabic meme safety detection. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Mohammed+Al%2DAni+and+Abdelrahman+Sameh+Ibrahim+Ibrahim+Hassaan%2E+2026%2E+DetectorAI+at+ArGuard+Shared+Tasks%3A+Calibrated+multimodal+and+label%2Dfrequency%2Daware+systems+for+Arabic+meme+safety+detection%2E+In+Pr)
5. Mansour Al Ghanim, Saleh Almohaimeed, Mengxin Zheng, Yan Solihin, and Qian Lou. 2024. Jailbreaking LLMs with Arabic transliteration and Arabizi. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 18584–18600, Miami, Florida, USA. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Mansour+Al+Ghanim%2C+Saleh+Almohaimeed%2C+Mengxin+Zheng%2C+Yan+Solihin%2C+and+Qian+Lou%2E+2024%2E+Jailbreaking+LLMs+with+Arabic+transliteration+and+Arabizi%2E+In+Proceedings+of+the+2024+Conference+on+Empirical+Meth)
6. Shahad Al-Khalifa, Nadir Durrani, Hend Al-Khalifa, and Firoj Alam. 2025. The landscape of arabic large language models. Communications of the ACM, 68(10):54–61. [find](https://scholar.google.com/scholar?q=Shahad+Al%2DKhalifa%2C+Nadir+Durrani%2C+Hend+Al%2DKhalifa%2C+and+Firoj+Alam%2E+2025%2E+The+landscape+of+arabic+large+language+models%2E+Communications+of+the+ACM%2C+68%2810%29%3A54%E2%80%9361%2E)
7. Firoj Alam, Stefano Cresci, Tanmoy Chakraborty, Fabrizio Silvestri, Dimiter Dimitrov, Giovanni Da San Martino, Shaden Shaar, Hamed Firooz, and Preslav Nakov. 2022. A survey on multimodal disinformation detection. In Proceedings of the 29th International Conference on Computational Linguistics, pages 6625–6643, Gyeongju, Republic of Korea. International Committee on Computational Linguistics. Firoj Alam, Abul Hasnat, Fatema Ahmad, Md. [find](https://scholar.google.com/scholar?q=Firoj+Alam%2C+Stefano+Cresci%2C+Tanmoy+Chakraborty%2C+Fabrizio+Silvestri%2C+Dimiter+Dimitrov%2C+Giovanni+Da+San+Martino%2C+Shaden+Shaar%2C+Hamed+Firooz%2C+and+Preslav+Nakov%2E+2022%2E+A+survey+on+multimodal+disinformatio)
8. Arid Hasan, and Maram Hasanain. 2024. ArMeme: Propagandistic content in Arabic memes. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 21071–21090, Miami, Florida, USA. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Arid+Hasan%2C+and+Maram+Hasanain%2E+2024%2E+ArMeme%3A+Propagandistic+content+in+Arabic+memes%2E+In+Proceedings+of+the+2024+Conference+on+Empirical+Methods+in+Natural+Language+Processing%2C+pages+21071%E2%80%9321090%2C+Miam)
9. Nouar AlDahoul and Yasir Zaki. 2026. NYUAD at Ar-Guard Shared Tasks: Multimodal Embedding Models for Detecting Arabic Hateful Memes and Unsafe Prompts. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Nouar+AlDahoul+and+Yasir+Zaki%2E+2026%2E+NYUAD+at+Ar%2DGuard+Shared+Tasks%3A+Multimodal+Embedding+Models+for+Detecting+Arabic+Hateful+Memes+and+Unsafe+Prompts%2E+In+Proceedings+of+the+Fourth+Arabic+Natural+Lang)
10. Yasser Ashraf, Yuxia Wang, Bin Gu, Preslav Nakov, and Timothy Baldwin. 2025. Arabic dataset for LLM safeguard evaluation. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), pages 5529–5546, Albuquerque, New Mexico. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Yasser+Ashraf%2C+Yuxia+Wang%2C+Bin+Gu%2C+Preslav+Nakov%2C+and+Timothy+Baldwin%2E+2025%2E+Arabic+dataset+for+LLM+safeguard+evaluation%2E+In+Proceedings+of+the+2025+Conference+of+the+Nations+of+the+Americas+Chapter+o)
11. Saoussen Ben Chaabane, Omar Trigui, and Maher Jaoua. 2026a. ANLP-UniSo at ArGuard 2026 Task A: Multimodal arabert–resnet50 with knowledge graph fusion for arabic hateful meme detection. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Saoussen+Ben+Chaabane%2C+Omar+Trigui%2C+and+Maher+Jaoua%2E+2026a%2E+ANLP%2DUniSo+at+ArGuard+2026+Task+A%3A+Multimodal+arabert%E2%80%93resnet50+with+knowledge+graph+fusion+for+arabic+hateful+meme+detection%2E+In+Proceedings)
12. Saoussen Ben Chaabane, Omar Trigui, and Maher Jaoua. 2026b. ANLP-UniSo at ArGuard 2026 Task B: Multi-task marbert with focal loss and hierarchical consistency for arabic harmful prompt detection. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Saoussen+Ben+Chaabane%2C+Omar+Trigui%2C+and+Maher+Jaoua%2E+2026b%2E+ANLP%2DUniSo+at+ArGuard+2026+Task+B%3A+Multi%2Dtask+marbert+with+focal+loss+and+hierarchical+consistency+for+arabic+harmful+prompt+detection%2E+In+P)
13. Yves Bestgen. 2026. SATLab at ArGuard Shared Tasks: Character n-grams for fine-grained categorization of harmful LLM prompts in Arabic. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Yves+Bestgen%2E+2026%2E+SATLab+at+ArGuard+Shared+Tasks%3A+Character+n%2Dgrams+for+fine%2Dgrained+categorization+of+harmful+LLM+prompts+in+Arabic%2E+In+Proceedings+of+the+Fourth+Arabic+Natural+Language+Processing+)
14. Dimitris Bilianos. 2026. dbilianos at ArGuard Shared Tasks: When does translation help in arabic harmful prompt detection? In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Dimitris+Bilianos%2E+2026%2E+dbilianos+at+ArGuard+Shared+Tasks%3A+When+does+translation+help+in+arabic+harmful+prompt+detection%3F+In+Proceedings+of+the+Fourth+Arabic+Natural+Language+Processing+Conference%3A+S)
15. Tiberiu Boros and Radu Chivereanu. 2026. RACAI at ArGuard Shared Task: Exploring self-taught LLM reasoners and fine-tuned transformers classifiers for Arabic prompt toxicity detection. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Tiberiu+Boros+and+Radu+Chivereanu%2E+2026%2E+RACAI+at+ArGuard+Shared+Task%3A+Exploring+self%2Dtaught+LLM+reasoners+and+fine%2Dtuned+transformers+classifiers+for+Arabic+prompt+toxicity+detection%2E+In+Proceedings+)
16. Minh Duc Bui, Katharina Von Der Wense, and Anne Lauscher. 2025. Multi 3 Hate: Multimodal, multilingual, and multicultural hate speech detection with vision–language models. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), pages 9714–9731. [find](https://scholar.google.com/scholar?q=Minh+Duc+Bui%2C+Katharina+Von+Der+Wense%2C+and+Anne+Lauscher%2E+2025%2E+Multi+3+Hate%3A+Multimodal%2C+multilingual%2C+and+multicultural+hate+speech+detection+with+vision%E2%80%93language+models%2E+In+Proceedings+of+the+2025+)
17. Rui Cao, Roy Ka-Wei Lee, Wen-Haw Chong, and Jing Jiang. 2022. Prompting for multimodal hateful meme classification. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, EMNLP ’22, pages 321 – 332. Association for Computational Linguistics (ACL). [find](https://scholar.google.com/scholar?q=Rui+Cao%2C+Roy+Ka%2DWei+Lee%2C+Wen%2DHaw+Chong%2C+and+Jing+Jiang%2E+2022%2E+Prompting+for+multimodal+hateful+meme+classification%2E+In+Proceedings+of+the+2022+Conference+on+Empirical+Methods+in+Natural+Language+Proce)
18. Dimitar Dimitrov, Firoj Alam, Maram Hasanain, Abul Hasnat, Fabrizio Silvestri, Preslav Nakov, and Giovanni Da San Martino. 2024. SemEval-2024 task 4: Multilingual detection of persuasion techniques in memes. In Proceedings of the 18th International Workshop on Semantic Evaluation (SemEval-2024), pages 2009–2026, Mexico City, Mexico. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Dimitar+Dimitrov%2C+Firoj+Alam%2C+Maram+Hasanain%2C+Abul+Hasnat%2C+Fabrizio+Silvestri%2C+Preslav+Nakov%2C+and+Giovanni+Da+San+Martino%2E+2024%2E+SemEval%2D2024+task+4%3A+Multilingual+detection+of+persuasion+techniques+in)
19. Abdelbasset Djamai, Khloud Al Jallad, Fatimah Emad Eldin, and Omer Nacar. 2026. NAMAA at ArGuard Shared Tasks: Multimodal Fusion, Vision–Language Model Adaptation, and In-Context Prompting for Arabic Hateful Meme Detection. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Abdelbasset+Djamai%2C+Khloud+Al+Jallad%2C+Fatimah+Emad+Eldin%2C+and+Omer+Nacar%2E+2026%2E+NAMAA+at+ArGuard+Shared+Tasks%3A+Multimodal+Fusion%2C+Vision%E2%80%93Language+Model+Adaptation%2C+and+In%2DContext+Prompting+for+Arabic+)
20. Masoomali Fatehkia, Enes Altinisik, and Husrev Taha Sencar. 2026. FanarGuard: A culturally-aware moderation filter for Arabic language models. In Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers), pages 7848–7869, Rabat, Morocco. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Masoomali+Fatehkia%2C+Enes+Altinisik%2C+and+Husrev+Taha+Sencar%2E+2026%2E+FanarGuard%3A+A+culturally%2Daware+moderation+filter+for+Arabic+language+models%2E+In+Proceedings+of+the+19th+Conference+of+the+European+Cha)
21. Elisabetta Fersini, Francesca Gasparini, Giulia Rizzi, Aurora Saibene, Berta Chulvi, Paolo Rosso, Alyssa Lees, and Jeffrey Sorensen. 2022. SemEval-2022 task 5: Multimedia automatic misogyny identification. In Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022), pages 533–549, Seattle, United States. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Elisabetta+Fersini%2C+Francesca+Gasparini%2C+Giulia+Rizzi%2C+Aurora+Saibene%2C+Berta+Chulvi%2C+Paolo+Rosso%2C+Alyssa+Lees%2C+and+Jeffrey+Sorensen%2E+2022%2E+SemEval%2D2022+task+5%3A+Multimedia+automatic+misogyny+identifica) [our card](https://md.research.elamur.ai/free/en/papers/fersini-2022-semeval-2022-task-5-multimedia/)
22. Ahmed M. Fetouh, Mariam Labib, Nsrin Ashraf, Omer Dawood, and Hamada Nayel. 2026. REGLAT at ArGuard Shared Tasks: An ensemble-based model integrating ner for harmful content detection in arabic. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Ahmed+M%2E+Fetouh%2C+Mariam+Labib%2C+Nsrin+Ashraf%2C+Omer+Dawood%2C+and+Hamada+Nayel%2E+2026%2E+REGLAT+at+ArGuard+Shared+Tasks%3A+An+ensemble%2Dbased+model+integrating+ner+for+harmful+content+detection+in+arabic%2E+In+Pr)
23. Seungju Han, Kavel Rao, Allyson Ettinger, Liwei Jiang, Bill Yuchen Lin, Nathan Lambert, Yejin Choi, and Nouha Dziri. 2024. WildGuard: Open one-stop moderation tools for safety risks, jailbreaks, and refusals of llms. Advances in neural information processing systems, 37:8093–8131. [find](https://scholar.google.com/scholar?q=Seungju+Han%2C+Kavel+Rao%2C+Allyson+Ettinger%2C+Liwei+Jiang%2C+Bill+Yuchen+Lin%2C+Nathan+Lambert%2C+Yejin+Choi%2C+and+Nouha+Dziri%2E+2024%2E+WildGuard%3A+Open+one%2Dstop+moderation+tools+for+safety+risks%2C+jailbreaks%2C+and+r) [our card](https://md.research.elamur.ai/free/en/papers/han-2024-wildguard-open-one-stop-moderation/)
24. Maram Hasanain, Firoj Alam, Hamdy Mubarak, Samir Abdaljalil, Wajdi Zaghouani, Preslav Nakov, Giovanni Da San Martino, and Abed Freihat. 2023. ArAIEval shared task: Persuasion techniques and disinformation detection in Arabic text. In Proceedings of ArabicNLP 2023, pages 483–493, Singapore (Hybrid). Association for Computational Linguistics. Maram Hasanain, Md. Arid Hasan, Fatema Ahmad, Reem Suwaileh, Md. [find](https://scholar.google.com/scholar?q=Maram+Hasanain%2C+Firoj+Alam%2C+Hamdy+Mubarak%2C+Samir+Abdaljalil%2C+Wajdi+Zaghouani%2C+Preslav+Nakov%2C+Giovanni+Da+San+Martino%2C+and+Abed+Freihat%2E+2023%2E+ArAIEval+shared+task%3A+Persuasion+techniques+and+disinforma)
25. Rafiul Biswas, Wajdi Zaghouani, and Firoj Alam. 2024. ArAIEval shared task: Propagandistic techniques detection in unimodal and multimodal Arabic content. In Proceedings of the Second Arabic Natural Language Processing Conference, pages 456–466, Bangkok, Thailand. Association for Computational Linguistics. Md. [find](https://scholar.google.com/scholar?q=Rafiul+Biswas%2C+Wajdi+Zaghouani%2C+and+Firoj+Alam%2E+2024%2E+ArAIEval+shared+task%3A+Propagandistic+techniques+detection+in+unimodal+and+multimodal+Arabic+content%2E+In+Proceedings+of+the+Second+Arabic+Natural+L)
26. Ajwad Hossain. 2026. Axiom at ArGuard Shared Tasks: Unimodal text dominance for arabic hate meme detection. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Ajwad+Hossain%2E+2026%2E+Axiom+at+ArGuard+Shared+Tasks%3A+Unimodal+text+dominance+for+arabic+hate+meme+detection%2E+In+Proceedings+of+the+Fourth+Arabic+Natural+Language+Processing+Conference%3A+Shared+Tasks%2C+Bu)
27. Emmanuel O. Ilesanmi, Olaronke Oluwayemisi Adebanji, Hamid T. Ogunlenu, Olumide Ebenezer Ojo, Hiram Calvo, and Alexander Gelbukh. 2026. TextLing at ArGuard Shared Tasks: An enhanced MARBERT-based late-fusion model for multimodal hateful meme detection in Arabic. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Emmanuel+O%2E+Ilesanmi%2C+Olaronke+Oluwayemisi+Adebanji%2C+Hamid+T%2E+Ogunlenu%2C+Olumide+Ebenezer+Ojo%2C+Hiram+Calvo%2C+and+Alexander+Gelbukh%2E+2026%2E+TextLing+at+ArGuard+Shared+Tasks%3A+An+enhanced+MARBERT%2Dbased+late)
28. Ali Khairallah and Arkaitz Zubiaga. 2026. AK_QMUL at ArGuard Shared Tasks: A single fine-tuned JAIS for Arabic harmful-prompt detection. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Ali+Khairallah+and+Arkaitz+Zubiaga%2E+2026%2E+AK%5FQMUL+at+ArGuard+Shared+Tasks%3A+A+single+fine%2Dtuned+JAIS+for+Arabic+harmful%2Dprompt+detection%2E+In+Proceedings+of+the+Fourth+Arabic+Natural+Language+Processing)
29. Selma Khelili and Zohra Abdeli. 2026. SZED at Ar-Guard Shared Tasks: Multi-task label-attention and cross-modal fusion for arabic hateful meme detection. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Selma+Khelili+and+Zohra+Abdeli%2E+2026%2E+SZED+at+Ar%2DGuard+Shared+Tasks%3A+Multi%2Dtask+label%2Dattention+and+cross%2Dmodal+fusion+for+arabic+hateful+meme+detection%2E+In+Proceedings+of+the+Fourth+Arabic+Natural+La)
30. Douwe Kiela, Hamed Firooz, Aravind Mohan, Vedanuj Goswami, Amanpreet Singh, Pratik Ringshia, and Davide Testuggine. 2020. The hateful memes challenge: Detecting hate speech in multimodal memes. In Advances in Neural Information Processing Systems, volume 33. [find](https://scholar.google.com/scholar?q=Douwe+Kiela%2C+Hamed+Firooz%2C+Aravind+Mohan%2C+Vedanuj+Goswami%2C+Amanpreet+Singh%2C+Pratik+Ringshia%2C+and+Davide+Testuggine%2E+2020%2E+The+hateful+memes+challenge%3A+Detecting+hate+speech+in+multimodal+memes%2E+In+Adv) [our card](https://md.research.elamur.ai/free/en/papers/kiela-2020-hateful-memes-challenge-detecting-hate/)
31. Mohamed Bayan Kmainasi, Abul Hasnat, Md Arid Hasan, Ali Ezzat Shahroor, and Firoj Alam. 2025. MemeIntel: Explainable detection of propagandistic and hateful memes. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 30251–30267, Suzhou, China. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Mohamed+Bayan+Kmainasi%2C+Abul+Hasnat%2C+Md+Arid+Hasan%2C+Ali+Ezzat+Shahroor%2C+and+Firoj+Alam%2E+2025%2E+MemeIntel%3A+Explainable+detection+of+propagandistic+and+hateful+memes%2E+In+Proceedings+of+the+2025+Conferenc)
32. Mohamed Bayan Kmainasi, Mucahid Kutlu, Ali Ezzat Shahroor, Abul Hasnat, and Firoj Alam. 2026a. Adapting reinforcement learning with chain-of-thought supervision for explainable detection of hateful and propagandistic memes. Machine Learning with Applications, page 101003. Mohamed Bayan Kmainasi, Ali Ezzat Shahroor, Abul Hasnat, Md. [find](https://scholar.google.com/scholar?q=Mohamed+Bayan+Kmainasi%2C+Mucahid+Kutlu%2C+Ali+Ezzat+Shahroor%2C+Abul+Hasnat%2C+and+Firoj+Alam%2E+2026a%2E+Adapting+reinforcement+learning+with+chain%2Dof%2Dthought+supervision+for+explainable+detection+of+hateful+an)
33. Rafiul Biswas, Wajdi Zaghouani, and Firoj Alam. 2026b. AHA-Memes: A fine-grained multimodal benchmark for understanding hate in arabic memes. arXiv preprint arXiv:2607.27393. [↗](https://arxiv.org/abs/2607.27393) [our card](https://md.research.elamur.ai/free/en/papers/kmainasi-2026-aha-memes-fine-grained-multimodal/)
34. Gokul Karthik Kumar and Karthik Nandakumar. 2022. Hate-CLIPper: Multimodal hateful meme classification based on cross-modal interaction of CLIP features. In Proceedings of the Second Workshop on NLP for Positive Impact (NLP4PI), pages 171–183, Abu Dhabi, United Arab Emirates (Hybrid). Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Gokul+Karthik+Kumar+and+Karthik+Nandakumar%2E+2022%2E+Hate%2DCLIPper%3A+Multimodal+hateful+meme+classification+based+on+cross%2Dmodal+interaction+of+CLIP+features%2E+In+Proceedings+of+the+Second+Workshop+on+NLP+f)
35. J Richard Landis and Gary G Koch. 1977. The measurement of observer agreement for categorical data. Biometrics, 33(1):159–174. [find](https://scholar.google.com/scholar?q=J+Richard+Landis+and+Gary+G+Koch%2E+1977%2E+The+measurement+of+observer+agreement+for+categorical+data%2E+Biometrics%2C+33%281%29%3A159%E2%80%93174%2E)
36. Ahmadullah Laskar and Sahinur Rahman Laskar. 2026. Bahash-AI at ArGuard Shared Tasks: Multimodal ensemble learning for hateful meme detection. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Ahmadullah+Laskar+and+Sahinur+Rahman+Laskar%2E+2026%2E+Bahash%2DAI+at+ArGuard+Shared+Tasks%3A+Multimodal+ensemble+learning+for+hateful+meme+detection%2E+In+Proceedings+of+the+Fourth+Arabic+Natural+Language+Proc)
37. Artiom Lipinski and Ilya Makarov. 2026. iMak AI Lab at ArGuard Shared Tasks: Heterogeneous lora ensembles of vision-language models for arabic hateful meme detection. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Artiom+Lipinski+and+Ilya+Makarov%2E+2026%2E+iMak+AI+Lab+at+ArGuard+Shared+Tasks%3A+Heterogeneous+lora+ensembles+of+vision%2Dlanguage+models+for+arabic+hateful+meme+detection%2E+In+Proceedings+of+the+Fourth+Arab)
38. Mantas Mazeika, Long Phan, Xuwang Yin, Andy Zou, Zifan Wang et al. 2024. Harmbench: a standardized evaluation framework for automated red teaming and robust refusal. In Proceedings of the 41st International Conference on Machine Learning, ICML’24. JMLR.org. [find](https://scholar.google.com/scholar?q=Mantas+Mazeika%2C+Long+Phan%2C+Xuwang+Yin%2C+Andy+Zou%2C+Zifan+Wang+et+al%2E+2024%2E+Harmbench%3A+a+standardized+evaluation+framework+for+automated+red+teaming+and+robust+refusal%2E+In+Proceedings+of+the+41st+Interna) [our card](https://md.research.elamur.ai/free/en/papers/mazeika-2024-harmbench-standardized-evaluation-framework-automated/)
39. Hamdy Mubarak, Hend Al-Khalifa, and Abdulmohsen Al-Thubaity. 2022. Overview of OSACT5 shared task on Arabic offensive language and hate speech detection. In Proceedings of the 5th Workshop on Open-Source Arabic Corpora and Processing Tools with Shared Tasks on Qur’an QA and Fine-Grained Hate Speech Detection, pages 162–166, Marseille, France. European Language Resources Association. [find](https://scholar.google.com/scholar?q=Hamdy+Mubarak%2C+Hend+Al%2DKhalifa%2C+and+Abdulmohsen+Al%2DThubaity%2E+2022%2E+Overview+of+OSACT5+shared+task+on+Arabic+offensive+language+and+hate+speech+detection%2E+In+Proceedings+of+the+5th+Workshop+on+Open%2DSou)
40. Hamdy Mubarak, Kareem Darwish, Walid Magdy, Tamer Elsayed, and Hend Al-Khalifa. 2020. Overview of OSACT4 Arabic offensive language detection shared task. In Proceedings of the 4th Workshop on Open-Source Arabic Corpora and Processing Tools, with a Shared Task on Offensive Language Detection, pages 48–52, Marseille, France. European Language Resources Association. [find](https://scholar.google.com/scholar?q=Hamdy+Mubarak%2C+Kareem+Darwish%2C+Walid+Magdy%2C+Tamer+Elsayed%2C+and+Hend+Al%2DKhalifa%2E+2020%2E+Overview+of+OSACT4+Arabic+offensive+language+detection+shared+task%2E+In+Proceedings+of+the+4th+Workshop+on+Open%2DSou)
41. Hamdy Mubarak, Abubakr Mohamed, and Majd Hawasly. 2025. AraSafe: Benchmarking safety in Arabic LLMs. In Findings of the Association for Computational Linguistics: EMNLP 2025, pages 9976–9992, Suzhou, China. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Hamdy+Mubarak%2C+Abubakr+Mohamed%2C+and+Majd+Hawasly%2E+2025%2E+AraSafe%3A+Benchmarking+safety+in+Arabic+LLMs%2E+In+Findings+of+the+Association+for+Computational+Linguistics%3A+EMNLP+2025%2C+pages+9976%E2%80%939992%2C+Suzhou%2C+)
42. Sidratul Muntaha, Sabila Anzum, Umme Saima Tonny, and Maimuna Tabassum. 2026. QuadCore at Ar-Guard Shared Tasks: Ensembling arabic encoders with dialect conditioning for harm-domain classification. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Sidratul+Muntaha%2C+Sabila+Anzum%2C+Umme+Saima+Tonny%2C+and+Maimuna+Tabassum%2E+2026%2E+QuadCore+at+Ar%2DGuard+Shared+Tasks%3A+Ensembling+arabic+encoders+with+dialect+conditioning+for+harm%2Ddomain+classification%2E+In)
43. Hamid T. Ogunlenu, Olaronke Oluwayemisi Adebanji, Emmanuel O. Ilesanmi, Olumide Ebenezer Ojo, Hiram Calvo, and Grigori Sidorov. 2026. CompLing at ArGuard Shared Tasks: Dialect-aware sparse fusion for Arabic prompt safety classification. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Hamid+T%2E+Ogunlenu%2C+Olaronke+Oluwayemisi+Adebanji%2C+Emmanuel+O%2E+Ilesanmi%2C+Olumide+Ebenezer+Ojo%2C+Hiram+Calvo%2C+and+Grigori+Sidorov%2E+2026%2E+CompLing+at+ArGuard+Shared+Tasks%3A+Dialect%2Daware+sparse+fusion+for+)
44. Olumide Ebenezer Ojo, Hamid T. Ogunlenu, Olaronke Oluwayemisi Adebanji, Tolulope Olalekan Abiola, Hiram Calvo, and Alexander Gelbukh. 2026. NLPLab at ArGuard Shared Tasks: Balancing confidence scores for fine-grained classification of unsafe Arabic prompts across harm domains. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Olumide+Ebenezer+Ojo%2C+Hamid+T%2E+Ogunlenu%2C+Olaronke+Oluwayemisi+Adebanji%2C+Tolulope+Olalekan+Abiola%2C+Hiram+Calvo%2C+and+Alexander+Gelbukh%2E+2026%2E+NLPLab+at+ArGuard+Shared+Tasks%3A+Balancing+confidence+scores+)
45. Miftahul Jannat Rishta and Sumaiya Zaman. 2026. Nile Nexus at ArGuard Shared Tasks: Leakage-controlled marbertv2–svm ensemble for arabic harmful-prompt classification. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Miftahul+Jannat+Rishta+and+Sumaiya+Zaman%2E+2026%2E+Nile+Nexus+at+ArGuard+Shared+Tasks%3A+Leakage%2Dcontrolled+marbertv2%E2%80%93svm+ensemble+for+arabic+harmful%2Dprompt+classification%2E+In+Proceedings+of+the+Fourth+Ara)
46. Wiam Saidi, Abdellatif El Abderrahmani, and Khalid Satori. 2026. L3ia morocco at arguard 2026: Intent over words - uncovering implicit harm in arabic llm prompts. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Wiam+Saidi%2C+Abdellatif+El+Abderrahmani%2C+and+Khalid+Satori%2E+2026%2E+L3ia+morocco+at+arguard+2026%3A+Intent+over+words+%2D+uncovering+implicit+harm+in+arabic+llm+prompts%2E+In+Proceedings+of+the+Fourth+Arabic+N)
47. Ali Ezzat Shahroor, Mohamed Bayan Kmainasi, Abul Hasnat, Dimitar Dimitrov, Giovanni Da San Martino, Preslav Nakov, and Firoj Alam. 2026. MemeLens: Multilingual multitask VLMs for memes. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, Budapest, Hungary. Association for Computational Linguistics. Shivam Sharma, Firoj Alam, Md. [find](https://scholar.google.com/scholar?q=Ali+Ezzat+Shahroor%2C+Mohamed+Bayan+Kmainasi%2C+Abul+Hasnat%2C+Dimitar+Dimitrov%2C+Giovanni+Da+San+Martino%2C+Preslav+Nakov%2C+and+Firoj+Alam%2E+2026%2E+MemeLens%3A+Multilingual+multitask+VLMs+for+memes%2E+In+Proceedings)
48. Shad Akhtar, Dimitar Dimitrov, Giovanni Da San Martino, Hamed Firooz, Alon Halevy, Fabrizio Silvestri, Preslav Nakov, and Tanmoy Chakraborty. 2022. Detecting and understanding harmful memes: A survey. In Proceedings of the Thirty-First International Joint Conference on Artificial Intelligence, IJCAI-22, pages 5597–5606. International Joint Conferences on Artificial Intelligence Organization. Survey Track. Md. [find](https://scholar.google.com/scholar?q=Shad+Akhtar%2C+Dimitar+Dimitrov%2C+Giovanni+Da+San+Martino%2C+Hamed+Firooz%2C+Alon+Halevy%2C+Fabrizio+Silvestri%2C+Preslav+Nakov%2C+and+Tanmoy+Chakraborty%2E+2022%2E+Detecting+and+understanding+harmful+memes%3A+A+survey%2E)
49. Faisal Sheikh. 2026. CalibGuard at ArGuard Shared Tasks: Threshold calibration and label-space constraints for Arabic multimodal hate speech and harmful prompt detection. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Faisal+Sheikh%2E+2026%2E+CalibGuard+at+ArGuard+Shared+Tasks%3A+Threshold+calibration+and+label%2Dspace+constraints+for+Arabic+multimodal+hate+speech+and+harmful+prompt+detection%2E+In+Proceedings+of+the+Fourth+)
50. Rahul Srivastava. 2026. Lattice AI at ArGuard Shared Tasks: Encoder ensembles and a re-annotated test slice in Arabic harmful-prompt detection. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Rahul+Srivastava%2E+2026%2E+Lattice+AI+at+ArGuard+Shared+Tasks%3A+Encoder+ensembles+and+a+re%2Dannotated+test+slice+in+Arabic+harmful%2Dprompt+detection%2E+In+Proceedings+of+the+Fourth+Arabic+Natural+Language+Pro)
51. Shardul Suryawanshi, Bharathi Raja Chakravarthi, Mihael Arcan, and Paul Buitelaar. 2020. Multimodal meme dataset (MultiOFF) for identifying offensive content in image and text. In Proceedings of the Second Workshop on Trolling, Aggression and Cyberbullying, pages 32–41, Marseille, France. European Language Resources Association (ELRA). [find](https://scholar.google.com/scholar?q=Shardul+Suryawanshi%2C+Bharathi+Raja+Chakravarthi%2C+Mihael+Arcan%2C+and+Paul+Buitelaar%2E+2020%2E+Multimodal+meme+dataset+%28MultiOFF%29+for+identifying+offensive+content+in+image+and+text%2E+In+Proceedings+of+the+S) [our card](https://md.research.elamur.ai/free/en/papers/suryawanshi-2020-multimodal-meme-dataset-multioff-identifying/)
52. Umme Saima Tonny, Maimuna Tabassum, Sidratul Muntaha, and Sabila Anzum. 2026. QuadCore at Ar-Guard Shared Tasks: A simple threshold calibration approach for arabic hateful meme detection. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Umme+Saima+Tonny%2C+Maimuna+Tabassum%2C+Sidratul+Muntaha%2C+and+Sabila+Anzum%2E+2026%2E+QuadCore+at+Ar%2DGuard+Shared+Tasks%3A+A+simple+threshold+calibration+approach+for+arabic+hateful+meme+detection%2E+In+Proceedin)
53. Sheikh Abdul Wahid, Muhammad Yamin, and Ali Arshad. 2026. Dynamos at ArGuard Shared Tasks: Multimodal transformers and selective ensembling for binary and fine-grained Arabic hateful meme and harmful prompt detection. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Sheikh+Abdul+Wahid%2C+Muhammad+Yamin%2C+and+Ali+Arshad%2E+2026%2E+Dynamos+at+ArGuard+Shared+Tasks%3A+Multimodal+transformers+and+selective+ensembling+for+binary+and+fine%2Dgrained+Arabic+hateful+meme+and+harmful+)
54. Zheng Xin Yong, Beyza Ermis, Marzieh Fadaee, Stephen Bach, and Julia Kreutzer. 2025. The state of multilingual LLM safety research: From measuring the language gap to mitigating it. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 15845–15860, Suzhou, China. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Zheng+Xin+Yong%2C+Beyza+Ermis%2C+Marzieh+Fadaee%2C+Stephen+Bach%2C+and+Julia+Kreutzer%2E+2025%2E+The+state+of+multilingual+LLM+safety+research%3A+From+measuring+the+language+gap+to+mitigating+it%2E+In+Proceedings+of+)
55. Ahmed Younis. 2026. Ahmed Younis at ArGuard 2026: Calibration is most of what you tune: Decode-time structure and threshold transport for Arabic harmful-content detection. In Proceedings of the Fourth Arabic Natural Language Processing Conference: Shared Tasks, Budapest, Hungary. Association for Computational Linguistics. Wajdi Zaghouani, Md. [find](https://scholar.google.com/scholar?q=Ahmed+Younis%2E+2026%2E+Ahmed+Younis+at+ArGuard+2026%3A+Calibration+is+most+of+what+you+tune%3A+Decode%2Dtime+structure+and+threshold+transport+for+Arabic+harmful%2Dcontent+detection%2E+In+Proceedings+of+the+Fourth)
56. Rafiul Biswas, Kholoud Khalil Aldous, and Mabrouka Bessghaier. 2026. Arabic-DialectSafety: A dialect-aware benchmark for arabic content safety classification. arXiv preprint arXiv:2608.01291. Wajdi Zaghouani, Md. [↗](https://arxiv.org/abs/2608.01291) [our card](https://md.research.elamur.ai/free/en/papers/zaghouani-2026-arabicdialectsafety-dialect-aware-benchmark-arabic/)
57. Rafiul Biswas, Mabrouka Bessghaier, Shimaa Ibrahim, George Mikros, Abul Hasnat, and Firoj Alam. 2025. MAHED shared task: Multimodal detection of hope and hate emotions in Arabic content. In Proceedings of The Third Arabic Natural Language Processing Conference: Shared Tasks, pages 560–574, Suzhou, China. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Rafiul+Biswas%2C+Mabrouka+Bessghaier%2C+Shimaa+Ibrahim%2C+George+Mikros%2C+Abul+Hasnat%2C+and+Firoj+Alam%2E+2025%2E+MAHED+shared+task%3A+Multimodal+detection+of+hope+and+hate+emotions+in+Arabic+content%2E+In+Proceeding)