Elamur Research⌕
☰
home / papers / 2026 / Fanar 2.0: Arabic Generative AI Stack

Fanar 2.0: Arabic Generative AI Stack

Paper · 2026 · FANAR TEAM, Ummar Abbas, Mohammad Shahmeer Ahmad et al.

Summary

Fanar 2.0 is QCRI's second-generation Arabic-centric generative AI platform. It is built and run entirely in-house, with no external AI providers, on a modest budget of 256 H100 GPUs. Arabic is also scarce online, at roughly 0.5% of web text. Instead of scaling up, the team chose careful curation. They continued pre-training Gemma-3-27B on about 120B selected tokens, split into three data recipes, and then applied recipe-based annealing and model merging. The resulting Fanar-27B model has a 32K context window and optional reasoning traces in Arabic and English. Although it used roughly eight times fewer tokens than Fanar 1.0, it improves on Arabic and English knowledge benchmarks and on dialect comprehension. Around the core LLM the platform adds:

  • a compact bilingual safety filter;
  • long-form speech recognition;
  • image generation and image/video understanding;
  • tool-calling agents and a multi-agent Islamic RAG system;
  • classical poetry generation and machine translation;
  • an orchestrator that routes requests by intent.

The work suggests that a sovereign, compute-limited effort can still compete with much larger projects.

Key points

  • Fanar-27B is trained by continual pre-training from Gemma-3-27B on ≈120B tokens, versus ≈1T tokens for Fanar 1.0, with all work done on 32 nodes × 8 H100.
  • Benchmark gains over Fanar 1.0:
    • MMMLU/Ar: +9.1
    • ArabicMMLU: +7.3
    • MMLU: +7.6
    • Belebele: +3.5
  • FanarGuard is a 4B moderation model trained on 468K Arabic/English prompt–response pairs, labelled for harmlessness and cultural alignment.
  • Aura-STT-LF handles hours-long recordings, including speaker changes and readability restoration, and comes with Aura-STT-BenchLF, a public Arabic long-form ASR benchmark.
  • The remaining components are:
    • Oryx-IG and Oryx-IVU (vision);
    • Fanar-Sadiq (multi-agent Islamic RAG);
    • Fanar-Diwan (poetry with joint diacritization);
    • FanarShaheen (translation);
    • an orchestrator with layered safety checks.

Interesting ideas

  • I1. Derived datasets are stored in folders named by a hash of the processing chain plus its run parameters, so any change in preprocessing automatically yields a separate, traceable training set.
  • I2. Intermediate artifacts are keyed by pipeline, task and run, and raw items are saved under their content hash, so a failed long job resumes from where it stopped and duplicates are caught early.
  • I3. Release is gated automatically: a retrained model is deployed only if its tracked metrics beat the model currently in production.
  • I4. A taxonomy tree of concepts generates targeted data queries per leaf, and user feedback is fed back to refine the tree and expose coverage gaps.
  • I5. Requests pass through several independent validation layers rather than a single filter.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • I1 (learn): Key each silver-data batch by a hash of the witness parsers, their versions and the agreement thresholds, so retrained models never silently mix data from different regimes.
  • I2 (learn): Apply the same layout to LLM-teacher annotation runs so that crashes do not waste paid calls and the same sentence is never labelled twice.
  • I3 (tree): Put the perceptron parser and gparse behind a LAS gate on held-out treebanks plus negative controls, so a regression can never replace the shipped parser.
  • I4 (world): Use the WordNet/VerbNet ontology as such a tree, generating test or harvest queries per verb class or concept so that thinly covered areas of world knowledge get filled deliberately.
  • I5 (answer): Check answers to word problems and fable QA at multiple stages (frame consistency, world-state check, final answer sanity) before release, with each layer's verdict kept for explanation.

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, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

🧭 Topics 1Research subjects and languages the work is about.Semitic languages · 2,908 works

Cite

Cite as FANAR TEAM, Ummar Abbas, Mohammad Shahmeer Ahmad, Minhaj Ahmad, Abdulaziz Al-Homaid, Anas Al-Nuaimi, Enes Altinisik, Ehsaneddin Asgari, Sanjay Chawla, Shammur Chowdhury, Fahim Dalvi, Kareem Darwish, Nadir Durrani, Mohamed Elfeky, Ahmed Elmagarmid, Mohamed Eltabakh, Asim Ersoy, Masoomali Fatehkia, Mohammed Qusay Hashim, Majd Hawasly, Mohamed Hefeeda, Mus'ab Husaini, Keivin Isufaj, Soon-Gyo Jung, Houssam Lachemat, Ji Kim Lucas, Abubakr Mohamed, Tasnim Mohiuddin, Basel Mousi, Hamdy Mubarak, Ahmad Musleh, Mourad Ouzzani, Amin Sadeghi, Husrev Taha Sencar, Mohammed Shinoy, Omar Sinan and Yifan Zhang. 2026. Fanar 2.0: Arabic Generative AI Stack. arXiv:2603.16397.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2603.16397 · arXiv 2603.16397

@misc{team2026fanar,
  title = {{Fanar 2.0: Arabic Generative AI Stack}},
  author = {FANAR TEAM, Ummar Abbas, Mohammad Shahmeer Ahmad, Minhaj Ahmad, Abdulaziz Al-Homaid, Anas Al-Nuaimi, Enes Altinisik, Ehsaneddin Asgari, Sanjay Chawla, Shammur Chowdhury, Fahim Dalvi, Kareem Darwish, Nadir Durrani, Mohamed Elfeky, Ahmed Elmagarmid, Mohamed Eltabakh, Asim Ersoy, Masoomali Fatehkia, Mohammed Qusay Hashim, Majd Hawasly, Mohamed Hefeeda, Mus'ab Husaini, Keivin Isufaj, Soon-Gyo Jung, Houssam Lachemat, Ji Kim Lucas, Abubakr Mohamed, Tasnim Mohiuddin, Basel Mousi, Hamdy Mubarak, Ahmad Musleh, Mourad Ouzzani, Amin Sadeghi, Husrev Taha Sencar, Mohammed Shinoy, Omar Sinan and Yifan Zhang},
  year = {2026},
  doi = {10.48550/arxiv.2603.16397},
  eprint = {2603.16397},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2603.16397}
}

References

100 references, 55 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Abbas, U., Ouzzani, M., Eltabakh, M.Y. et al (2026). Fanar-sadiq: A multi-agent architecture for grounded islamic qa. arXiv preprint arXiv:2603.08501. ↗
  2. Abdelali, A., Mubarak, H., Chowdhury, S.A. et al (2024). LAraBench: Benchmarking Arabic AI with Large Language Models. In Graham, Y. and Purver, M., editors, Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers), pages 487–520, St. Julian’s, Malta. Association for Computational Linguistics. ↗
  3. Al-Khalifa, S., Durrani, N., Al-Khalifa, H. et al (2025). The landscape of arabic large language models (allms): A new era for arabic language technology. Communications of the ACM. Online First. ↗
  4. Alam, F., Shahroor, A.E., Hasan, M.A. et al (2025). Everydaymmqa: A multilingual and multimodal framework for culturally grounded spoken visual qa. find
  5. Ali, A., Bell, P., Glass, J. et al (2016). The MGB-2 challenge: Arabic multi-dialect broadcast media recognition. In 2016 IEEE Spoken Language Technology Workshop (SLT), pages 279–284. IEEE.
  6. Ali, A., Chowdhury, S., Hussein, A. et al (2021). Arabic Code-Switching Speech Recognition using Monolingual Data. In Proc. of the 22nd Annual Conference of the International Speech Communication Association (INTERSPEECH), pages 3475–3479, Brno, Czechia.
  7. Altakrori, M.H., Habash, N., Freihat, A. et al (2025). DialectalArabicMMLU: Benchmarking dialectal capabilities in Arabic and multilingual language models.
  8. Altinisik, E., Fatehkia, M., Deniz, F. et al (2026). Do i really know? learning factual self-verification for hallucination reduction. find
  9. Alwajih, F., El Mekki, A., Magdy, S.M. et al (2025a). Palm: A culturally inclusive and linguistically diverse dataset for Arabic LLMs. In Che, W., Nabende, J., Shutova, E. et al, editors, Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 32871–32894, Vienna, Austria. Association for Computational Linguistics. ↗
  10. Alwajih, F., El Mekki, A., Mubarak, H. et al (2025b). PalmX 2025: The first shared task on benchmarking LLMs on Arabic and Islamic culture. In Proceedings of The Third Arabic Natural Language Processing Conference: Shared Tasks, pages 774–789.
  11. AMC (2023). American mathematics competitions 2023. find
  12. Antoun, W., Baly, F. and Hajj, H. (2021). AraGPT2: Pre-trained transformer for Arabic language generation. In Habash, N., Bouamor, H., Hajj, H. et al, editors, Proceedings of the Sixth Arabic Natural Language Processing Workshop, pages 196–207, Kyiv, Ukraine (Virtual). Association for Computational Linguistics. ↗
  13. Anwar, M., Freihat, A., Ibrahim, G. et al (2025). Jais 2: A family of Arabic-centric open large language models. Technical report, IFM. ↗
  14. Applied Innovation Center (2026). Karnak Model Applied-Innovation-Center/Karnak. Accessed on Mar 9, 2026. Card. https://huggingface.co/ ↗
  15. Bai, S., Chen, K., Liu, X. et al (2025). Qwen2. 5-vl technical report. arXiv preprint arXiv:2502.13923. — Qwen2.5-VL Technical Report ↗
  16. Bandarkar, L., Liang, D., Muller, B. et al (2024). The belebele benchmark: a parallel reading comprehension dataset in 122 language variants. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 749–775, Bangkok, Thailand and virtual meeting. Association for Computational Linguistics.
  17. Bari, M.S., Alnumay, Y., Alzahrani, N.A. et al (2024). ALLaM: Large Language Models for Arabic and English. arXiv preprint arXiv:2407.15390. ↗
  18. Bhargava, A., Witkowski, C., Looi, S.Z. et al (2023). What’s the magic word? a control theory of llm prompting. arXiv preprint arXiv:2310.04444. ↗
  19. Biderman, S., Schoelkopf, H., Anthony, Q. et al (2023). Pythia: a suite for analyzing large language models across training and scaling. In Proceedings of the 40th International Conference on Machine Learning, ICML’23. JMLR.org. find
  20. Bisk, Y., Zellers, R., Gao, J. et al (2020). PIQA: Reasoning about physical commonsense in natural language. In Proceedings of the AAAI conference on artificial intelligence, volume 34, pages 7432–7439. ↗
  21. Boughorbel, S. and Hawasly, M. (2023). Analyzing multilingual competency of LLMs in multi-turn instruction following: A case study of Arabic. In Sawaf, H., El-Beltagy, S., Zaghouani, W. et al, editors, Proceedings of ArabicNLP 2023, pages 128–139, Singapore (Hybrid). Association for Computational Linguistics.
  22. Bredin, H. (2023). pyannote.audio 2.1 speaker diarization pipeline: principle, benchmark, and recipe. In Proc. INTERSPEECH 2023. find
  23. Chen, G., Chai, S., Wang, G. et al (2021). Gigaspeech: An evolving, multi-domain ASR corpus with 10,000 hours of transcribed audio. In Proc. Interspeech 2021. find
  24. Chen, J., Zhang, T., Liu, C. et al (2025a). Taskgalaxy: Scaling multi-modal instruction fine-tuning with tens of thousands vision task types. In Yue, Y., Garg, A., Peng, N. et al, editors, International Conference on Learning Representations, volume 2025, pages 92969–93019. ↗
  25. Chen, Y., Niu, Z., Ma, Z. et al (2025b). F5-TTS: A fairytaler that fakes fluent and faithful speech with flow matching. In Che, W., Nabende, J., Shutova, E. et al, editors, Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 6255–6271, Vienna, Austria. Association for Computational Linguistics.
  26. Clark, P., Cowhey, I., Etzioni, O. et al (2018). Think you have solved question answering? try arc, the ai2 reasoning challenge. arXiv preprint arXiv:1803.05457. ↗
  27. Cobbe, K., Kosaraju, V., Bavarian, M. et al (2021). Training verifiers to solve math word problems. arXiv preprint arXiv:2110.14168. ↗
  28. Conneau, A., Khandelwal, K., Goyal, N. et al (2020). Unsupervised cross-lingual representation learning at scale. In Jurafsky, D., Chai, J., Schluter, N. et al, editors, Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 8440–8451, Online. Association for Computational Linguistics. find
  29. Deitke, M., Clark, C., Lee, S. et al (2025). Molmo and pixmo: Open weights and open data for state-of-the-art vision-language models. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages 91–104.
  30. Deniz, F., Popovic, D., Boshmaf, Y. et al (2025). aixamine: Simplified llm safety and security. arXiv preprint arXiv:2504.14985. ↗
  31. Ding, N., Chen, Y., Xu, B. et al (2023). Enhancing chat language models by scaling high-quality instructional conversations. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 3029–3051. ↗
  32. Doan, K., Waheed, A. and Abdul-Mageed, M. (2024). Towards zero-shot text-to-speech for Arabic dialects. In Proceedings of the Second Arabic Natural Language Processing Conference, pages 123–129.
  33. El Filali, A., ALOUI, M., Husaain, T. et al (2025). Open arabic llm leaderboard 2. https://huggingface.co/spaces/OALL/Open-Arabic-LLM-Leaderboard. ↗
  34. Elfilali, A., Alobeidli, H., Fourrier, C. et al (2024). Open arabic llm leaderboard. https://huggingface.co/ spaces/OALL/Open-Arabic-LLM-Leaderboard. ↗
  35. Ersoy, A., Altinisik, E., Darwish, K.M. et al (2025). Tool calling for Arabic llms: Data strategies and instruction tuning. In Proceedings of The Third Arabic Natural Language Processing Conference, pages 347– 358.
  36. Fanar Team, Abbas, U., Ahmad, M.S. et al (2025). Fanar: An Arabic-Centric Multimodal Generative AI Platform. arxXiv:2501.13944.
  37. Fatehkia, M., Altinisik, E. and Sencar, H.T. (2025). FanarGuard: A culturally-aware moderation filter for arabic language models. arXiv preprint arXiv:2511.18852. ↗
  38. Fox, J.D., Raj, D., Delworth, N. et al (2024). Updated corpora and benchmarks for long-form speech recognition. In ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 13246–13250. find
  39. Gemini Team, Anil, R., Borgeaud, S. et al (2023). Gemini: a family of highly capable multimodal models. arXiv preprint arXiv:2312.11805. ↗
  40. Gemma Team (2024). Gemma: Open models based on gemini research and technology. arXiv preprint arXiv:2403.08295. ↗
  41. Ghaboura, S., Heakl, A., Thawakar, O. et al (2025). Camel-bench: A comprehensive Arabic lmm benchmark. In Findings of the Association for Computational Linguistics: NAACL 2025, pages 1970–1980. ↗
  42. Goddard, C., Siriwardhana, S., Ehghaghi, M. et al (2024). Arcee’s MergeKit: A toolkit for merging large language models. In Dernoncourt, F., Preoţiuc-Pietro, D. and Shimorina, A., editors, Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: Industry Track, pages 477–485, Miami, Florida, US. Association for Computational Linguistics. ↗
  43. Gunasekar, S., Zhang, Y., Aneja, J. et al (2023). Textbooks are all you need. ↗
  44. Gupta, K., Thérien, B., Ibrahim, A. et al (2023). Continual pre-training of large language models: How to (re)warm your model? ↗
  45. Gururangan, S., Marasović, A., Swayamdipta, S. et al (2020). Don’t stop pretraining: Adapt language models to domains and tasks. In Jurafsky, D., Chai, J., Schluter, N. et al, editors, Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 8342–8360, Online. Association for Computational Linguistics. find
  46. Habash, N.Y. (2010). Introduction to Arabic natural language processing. Morgan & Claypool Publishers. ↗
  47. Harper, J. and NVIDIA NeMo Team (2019–2026). NeMo: a toolkit for building AI applications using Neural Modules. https://github.com/NVIDIA/NeMo. ↗
  48. Hawasly, M., Mohiuddin, T., Mubarak, H. et al (2025). ArabicWeb-edu: Educational quality data for Arabic LLM training. In Darwish, K., Ali, A., Abu Farha, I. et al, editors, Proceedings of The Third Arabic Natural Language Processing Conference, pages 436–447, Suzhou, China. Association for Computational Linguistics.
  49. Hendrycks, D., Burns, C., Basart, S. et al (2020). Measuring massive multitask language understanding. In International Conference on Learning Representations. ↗
  50. Hoffmann, J., Borgeaud, S., Mensch, A. et al (2022). Training compute-optimal large language models. In Advances in Neural Information Processing Systems 35 (NeurIPS 2022), NIPS ’22, Red Hook, NY, USA. Curran Associates Inc.
  51. Huang, H., Yu, F., Zhu, J. et al (2024). AceGPT, localizing large language models in Arabic. In Duh, K., Gomez, H. and Bethard, S., editors, Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), pages 8139–8163, Mexico City, Mexico. Association for Computational Linguistics. ↗
  52. Ilharco, G., Ribeiro, M.T., Wortsman, M. et al (2023). Editing models with task arithmetic. In The Eleventh International Conference on Learning Representations. ↗
  53. Ito, K. and Johnson, L. (2017). The lj speech dataset. https://keithito.com/LJ-Speech-Dataset/. ↗
  54. Kaplan, J., McCandlish, S., Henighan, T. et al (2020). Scaling laws for neural language models. ↗
  55. Koto, F., Li, H., Shatnawi, S. et al (2024). ArabicMMLU: Assessing massive multitask language understanding in Arabic. In Ku, L.W., Martins, A. and Srikumar, V., editors, Findings of the Association for Computational Linguistics: ACL 2024, pages 5622–5640, Bangkok, Thailand. Association for Computational Linguistics. ↗
  56. Li, X., Takamichi, S., Saeki, T. et al (2023a). Yodas: Youtube-oriented dataset for audio and speech. In 2023 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU), pages 1–8. IEEE. find
  57. Li, Y., Bubeck, S., Eldan, R. et al (2023b). Textbooks are all you need ii: phi-1.5 technical report. ↗
  58. Liang, Q., Liu, Z., Ostrow, M. et al (2024). How diffusion models learn to factorize and compose. In Globerson, A., Mackey, L., Belgrave, D. et al, editors, Advances in neural information processing systems, volume 37, pages 15121–15148. Curran Associates, Inc. find
  59. Lightman, H., Kosaraju, V., Burda, Y. et al (2024). Let’s verify step by step. In The Twelfth International Conference on Learning Representations. ↗
  60. Liu, Z., Hoang, T.Q., Zhang, J. et al (2024). APIGen: Automated PIpeline for generating verifiable and diverse function-calling datasets. In The Thirty-eight Conference on Neural Information Processing Systems Datasets and Benchmarks Track. find
  61. Lozhkov, A., Ben Allal, L., von Werra, L. et al (2024). Fineweb-edu. https://huggingface.co/datasets/ HuggingFaceFW/fineweb-edu. ↗
  62. Luo, Y., Yang, Z., Meng, F. et al (2025). An empirical study of catastrophic forgetting in large language models during continual fine-tuning. ↗
  63. Masterman, T., Besen, S., Sawtell, M. et al (2024). The landscape of emerging ai agent architectures for reasoning, planning, and tool calling: A survey. arXiv preprint arXiv:2404.11584. ↗
  64. Meta AI (2024). The llama 3 herd of models. arXiv preprint arXiv:2407.21783. ↗
  65. Mohamed, A. and Mubarak, H. (2025). Advancing Arabic diacritization: Improved datasets, benchmarking, and state-of-the-art models. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 16718–16730. ↗
  66. Mousi, B., Dalvi, F., Chowdhury, S. et al (2026). Once correct, still wrong: Counterfactual hallucination in multilingual vision-language models. ↗
  67. Mousi, B., Durrani, N., Ahmad, F. et al (2025). AraDiCE: Benchmarks for dialectal and cultural capabilities in LLMs. In Rambow, O., Wanner, L., Apidianaki, M. et al, editors, Proceedings of the 31st International Conference on Computational Linguistics, pages 4186–4218, Abu Dhabi, UAE. Association for Computational Linguistics. ↗
  68. Mubarak, H. and Darwish, K. (2014). Automatic correction of Arabic text: A cascaded approach. In Habash, N. and Vogel, S., editors, Proceedings of the EMNLP 2014 workshop on Arabic natural language processing (ANLP), pages 132–136, Doha, Qatar. Association for Computational Linguistics.
  69. Mubarak, H., Hawasly, M. and Mohamed, A. (2026). Nahw: A comprehensive benchmark of Arabic grammar understanding, error detection, correction, and explanation. In 19th Conference of the European Chapter of the Association for Computational Linguistics.
  70. Mubarak, H., Hussein, A., Chowdhury, S.A. et al (2021). QASR: QCRI Aljazeera Speech Resource. A Large Scale Annotated Arabic Speech Corpus. In Zong, C., Xia, F., Li, W. et al, editors, Proc. of the 59th Annual Meeting of the Association for Computational Linguistics (ACL), pages 2274–2285, Online. Association for Computational Linguistics.
  71. Musleh, A., Zhang, Y. and Darwish, K. (2026). More data, fewer diacritics: Scaling Arabic TTS. ↗
  72. NLLB Team, Costa-jussà, M.R., Cross, J. et al (2022). No language left behind: Scaling human-centered machine translation. arXiv preprint arXiv:2207.04672. ↗
  73. OpenAI (2024). Multilingual massive multitask language understanding. datasets/openai/MMMLU. https://huggingface.co/ ↗
  74. Ostris (2025). ai-toolkit: The ultimate training toolkit for finetuning diffusion models. https://github.com/ostris/ai-toolkit. Accessed: 2026-02-26. ↗
  75. Panayotov, V., Chen, G., Povey, D. et al (2015). Librispeech: an ASR corpus based on public domain audio books. In 2015 IEEE international conference on acoustics, speech and signal processing (ICASSP), pages 5206–5210. IEEE. find
  76. Park, S.J., Salazar, J., Jansen, A. et al (2024). Long-form speech generation with spoken language models. CoRR, abs/2412.18603. ↗
  77. Peng, Y., Sudo, Y., Shakeel, M. et al (2024). OWSM-CTC: An open encoder-only speech foundation model for speech recognition, translation, and language identification. In Ku, L.W., Martins, A. and Srikumar, V., editors, Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 10192–10209, Bangkok, Thailand. Association for Computational Linguistics. find
  78. Perez, E., Huang, S., Song, F. et al (2022). Red teaming language models with language models. In Goldberg, Y., Kozareva, Z. and Zhang, Y., editors, Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 3419–3448, Abu Dhabi, United Arab Emirates. Association for Computational Linguistics. ↗
  79. Plaquet, A. and Bredin, H. (2023). Powerset multi-class cross entropy loss for neural speaker diarization. In Proc. INTERSPEECH 2023. find
  80. Qwen Team (2025). Qwen3 technical report. find
  81. Rana, A. (2010). Common crawl–building an open web-scale crawl using hadoop. https://commoncrawl.org/blog/slideshare-building-a-scalable-web-crawler-with-hadoop. Accessed on Dec 1, 2024. ↗
  82. Sajjad, H., Abdelali, A., Durrani, N. et al (2020). AraBench: Benchmarking dialectal Arabic-English machine translation. In Proceedings of the 28th International Conference on Computational Linguistics, pages 5094– 5107, Barcelona, Spain (Online). International Committee on Computational Linguistics. find
  83. Silero Team (2024). Silero vad: pre-trained enterprise-grade voice activity detector (vad), number detector and language classifier. https://github.com/snakers4/silero-vad. ↗
  84. Siuzdak, H. (2024). Vocos: Closing the gap between time-domain and fourier-based neural vocoders for high-quality audio synthesis. In Kim, B., Yue, Y., Chaudhuri, S. et al, editors, International Conference on Learning Representations, volume 2024, pages 25719–25733. ↗
  85. Sukhadia, V.N. and Chowdhury, S.A. (2025). HARNESS: Lightweight distilled Arabic speech foundation models.
  86. Sukhadia, V.N., Lachemat, H.E.O., Bhatti, H.H. et al (2026). Toward robust long-context arabic ASR: A real-world benchmark. Technical report, Qatar Computing Research Institute, Doha, Qatar. find
  87. Tiedemann, J. and Thottingal, S. (2020). OPUS-MT – building open translation services for the world. In Proceedings of the 22nd Annual Conference of the European Association for Machine Translation, pages 479–480, Lisboa, Portugal. European Association for Machine Translation. find
  88. Veaux, C., Yamagishi, J. and MacDonald, K. (2017). CSTR VCTK Corpus: English multi-speaker corpus for CSTR voice cloning toolkit. find
  89. Wei, A., Haghtalab, N. and Steinhardt, J. (2023). Jailbroken: How does llm safety training fail? Advances in Neural Information Processing Systems, 36:80079–80110.
  90. Wei, J., Tay, Y., Bommasani, R. et al (2022). Emergent abilities of large language models. Transactions on Machine Learning Research.
  91. Wolf, Y., Wies, N., Avnery, O. et al (2024). Fundamental limitations of alignment in large language models. In Proceedings of the 41st International Conference on Machine Learning, ICML’24. JMLR.org. ↗
  92. Wortsman, M., Ilharco, G., Gadre, S.Y. et al (2022). Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time. In Chaudhuri, K., Jegelka, S., Song, L. et al, editors, Proceedings of the 39th International Conference on Machine Learning, volume 162 of Proceedings of Machine Learning Research, pages 23965–23998. PMLR. ↗
  93. Yadav, P., Tam, D., Choshen, L. et al (2023). TIES-merging: Resolving interference when merging models. In Proceedings of the 37th International Conference on Neural Information Processing Systems.
  94. Zbib, M., Hammoud, H.A.A.K., Mukalled, S. et al (2025). AraLingBench: A human-annotated benchmark for evaluating arabic linguistic capabilities of large language models. arXiv preprint arXiv:2511.14295. ↗
  95. Zellers, R., Holtzman, A., Bisk, Y. et al (2019). HellaSwag: Can a machine really finish your sentence? In Korhonen, A., Traum, D. and Màrquez, L., editors, Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 4791–4800, Florence, Italy. Association for Computational Linguistics. ↗
  96. Zhang, Y. and Math-AI Team (2024). American Invitational Mathematics Examination (AIME) 2024. find
  97. Zheng, L., Chiang, W.L., Sheng, Y. et al (2023). Judging LLM-as-a-judge with MT-bench and chatbot arena. In Thirty-seventh Conference on Neural Information Processing Systems Datasets and Benchmarks Track.
  98. Zheng, Y., Zhang, R., Zhang, J. et al (2024). LlamaFactory: Unified efficient fine-tuning of 100+ language models. In ACL. Association for Computational Linguistics.
  99. Zhou, J., Lu, T., Mishra, S. et al (2023). Instruction-following evaluation for large language models. ↗
  100. Zhu, J., Huang, H., Lin, Z. et al (2025). Second language (Arabic) acquisition of LLMs via progressive vocabulary expansion. In Che, W., Nabende, J., Shutova, E. et al, editors, Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 2025–2042, Vienna, Austria. Association for Computational Linguistics. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/team-2026-fanar-2-0-arabic-generative/