---
title: "mucAI at Ahasis Shared Task — paper by Abdou (2025)"
description: "The author compares GPT-4o under three prompting regimes: no examples, a fixed hand-picked set of demonstrations, and… 18 references."
authors: ["Ahmed Mohamed Abdelaal Abdou"]
year: 2025
venue: "Proceedings of the Shared Task on Sentiment Analysis for Arabic Dialects"
source: "https://aclanthology.org/2025.ranlp-ahasis.9/"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/abdou-2025-mucai-ahasis-shared-task-sentiment/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/abdou-2025-mucai-ahasis-shared-task-sentiment/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2025](https://md.research.elamur.ai/free/en/papers/2025/page/18/) / mucAI at Ahasis Shared Task: Sentiment Analysis with Adaptive Few Shot Prompting

# mucAI at Ahasis Shared Task: Sentiment Analysis with Adaptive Few Shot Prompting

**Authors:** Ahmed Mohamed Abdelaal Abdou · **Year:** 2025 · **Venue:** [Proceedings of the Shared Task on Sentiment Analysis for Arabic Dialects](https://md.research.elamur.ai/free/en/venues/ranlp/) ([RANLP 2025](https://md.research.elamur.ai/free/en/venues/ranlp/2025/)) · **Publisher:** [INCOMA Ltd.](https://md.research.elamur.ai/free/en/publishers/incoma-ltd/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2025.ranlp-ahasis.9/) · [PDF](https://aclanthology.org/2025.ranlp-ahasis.9.pdf) · [OpenAlex](https://openalex.org/W7140089685)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
The paper tackles three-way sentiment classification (positive, negative, neutral) of hotel reviews written in two Arabic dialects, Saudi and Darija, for the AHASIS 2025 shared task. The author compares GPT-4o under three prompting regimes: no examples, a fixed hand-picked set of demonstrations, and an adaptive scheme. Fine-tuned AraBERT and MARBERT serve as baselines. In the adaptive scheme, each incoming review is embedded with an AraBERT encoder. Similar training reviews are then retrieved and balanced across sentiment labels. Each retrieved example comes with a GPT-4o explanation that was written with knowledge of its correct label. These explanations are produced once and cached.

The adaptive variant scored best on the hidden test set with 76.0 macro-F1, which placed sixth on the leaderboard. Both encoder baselines reached 73 on test. The main gain came on the neutral class, the hardest one. The work shows that choosing demonstrations per query can make an LLM adopt a dataset's own notion of a label, without any fine-tuning.

## Key points
- Data: 860 training reviews (430 per dialect; per dialect 154 positive, 168 negative, 108 neutral), split 80/20 into 688 train and 172 dev; the test set has 216 reviews without sentiment labels.
- Test macro-F1 by system:

  | System | Test macro-F1 |
  |---|---|
  | AraBERT | 73 |
  | MARBERT | 73 |
  | Zero-shot GPT-4o | 75 |
  | Static few-shot | 74 |
  | Adaptive few-shot | 76 |

  MARBERT had the best dev score (85.2) but fell to 73 on test.
- Retrieval settings: k = 20 nearest neighbours, then at most 3 per sentiment class, which gives 3–9 demonstrations. Each demonstration is a (review, label, rationale) triple.
- Neutral-class F1 on the dev set:
  - Saudi: 70 (zero-shot), 78 (static), 81 (adaptive).
  - Darija: 63, 62 and 67 for the same three strategies.
  - Darija lagged Saudi overall (macro 80 vs 89 with adaptive prompting).
- Cost per test item: about 778 input and 239 output tokens for the main call. On top of that come 3–9 cached rationale calls of roughly 95 input and 219 output tokens each.

## Interesting ideas
- Have the teacher model explain an example while already knowing its correct label, and cache that rationale so each training item is explained only once. This yields cheap, reusable worked examples.
- Retrieve many neighbours, then cap how many come from each label, so demonstrations stay relevant without letting the majority label dominate the prompt.
- The error analysis found that zero-shot GPT-4o applied a general-language idea of "neutral" rather than the dataset's narrower definition, and nearby annotated cases corrected this. Local examples act as an operational definition of a fuzzy category.
- The best dev model (MARBERT) lost its lead on test, a warning against trusting a small dev split carved from the training data.

## For Mova

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

- **I1** (learn): LLM teacher tiers could attach gold-conditioned explanations to silver parses or frame annotations once, building a reusable bank of justified examples for later prompting or rule induction.
- **I2** (learn): when choosing examples to show a teacher LLM for hard calls such as dependency labels or semantic roles, per-label quotas could curb the teacher's drift toward frequent labels.
- **I3** (frames): fuzzy pragmatic categories such as irony, implicature or indirect speech acts could be grounded by showing the closest annotated cases, rather than relying on a generic definition.
- **I4** (learn): evaluation should keep a truly held-out set alongside negative controls, so that a parser or reader chosen for its dev score is not overestimated.

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

## Institutions

- [Technical University of Munich](https://md.research.elamur.ai/free/en/universities/technical-university-of-munich/) — University, Munich, Germany · 62 works · 48,970 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works

🧭 Topics (1): [😊 Sentiment & emotion](https://md.research.elamur.ai/free/en/topics/sentiment-emotion/) · 1,035 works

## Cite

Cite as: Ahmed Mohamed Abdelaal Abdou. 2025. mucAI at Ahasis Shared Task: Sentiment Analysis with Adaptive Few Shot Prompting. Proceedings of the Shared Task on Sentiment Analysis for Arabic Dialects.

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

```bibtex
@inproceedings{abdou2025mucai,
  title = {{mucAI at Ahasis Shared Task: Sentiment Analysis with Adaptive Few Shot Prompting}},
  author = {Ahmed Mohamed Abdelaal Abdou},
  booktitle = {Proceedings of the Shared Task on Sentiment Analysis for Arabic Dialects},
  year = {2025},
  doi = {10.26615/978-954-452-109-7-009},
  url = {https://doi.org/10.26615/978-954-452-109-7-009}
}
```

- [← previous Mubeen AI: A Specialized Arabic Language Model for Heritage Preservation and User Intent Understanding](https://md.research.elamur.ai/free/en/papers/aljafari-2025-mubeen-ai-specialized-arabic-language/)
- [next → MuDRiC: Multi-Dialect Reasoning for Arabic Commonsense Validation](https://md.research.elamur.ai/free/en/papers/elozeiri-2025-mudric-multi-dialect-reasoning-arabic/)
- [↑ back to papers of 2025](https://md.research.elamur.ai/free/en/papers/2025/page/18/)

## References

18 references, 14 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Muhammad Abdul-Mageed, AbdelRahim Elmadany, and El Moatez Billah Nagoudi. 2021. [ARBERT & MARBERT: Deep bidirectional transformers for arabic](https://md.research.elamur.ai/free/en/papers/abdul-mageed-2021-arbert-marbert-deep-bidirectional-transformers/). In Proc. ACL, pages 7088–7105.
2. Abdulmohsen Al-Thubaity, Sakhar Alkhereyf, Hanan Murayshid, Nouf Alshalawi, Maha Omirah, Raghad Alateeq, Rawabi Almutairi, Razan Alsuwailem, Manal Alhassoun, and Imaan Alkhanen. 2023. [Evaluating ChatGPT and bard AI on Arabic sentiment analysis](https://md.research.elamur.ai/free/en/papers/thubaity-2023-evaluating-chatgpt-bard-ai-arabic/). In Proceedings of ArabicNLP 2023, pages 335–349, Singapore (Hybrid). Association for Computational Linguistics.
3. Maram Alharbi, Salmane Chafik, Saad Ezzini, Ruslan Mitkov, Tharindu Ranasinghe, and Hansi Hettiarachchi. 2025a. [Ahasis: Shared task on sentiment analysis for arabic dialects](https://md.research.elamur.ai/free/en/papers/alharbi-2025-ahasis-shared-task-sentiment-analysis/). In Proceedings of the 15th International Conference on Recent Advances in Natural Language Processing (RANLP).
4. Maram Alharbi, Saad Ezzini, Hansi Hettiarachchi, Tharindu Ranasinghe, and Ruslan Mitkov. 2025b. Evaluating large language models on arabic dialect sentiment analysis. In Proceedings of the 15th International Conference on Recent Advances in Natural Language Processing (RANLP). [find](https://scholar.google.com/scholar?q=Maram+Alharbi%2C+Saad+Ezzini%2C+Hansi+Hettiarachchi%2C+Tharindu+Ranasinghe%2C+and+Ruslan+Mitkov%2E+2025b%2E+Evaluating+large+language+models+on+arabic+dialect+sentiment+analysis%2E+In+Proceedings+of+the+15th+Intern)
5. Zaid Alyafeai, Maged S. Alshaibani, Badr AlKhamissi, Hamzah Luqman, Ebrahim Alareqi, and Ali Fadel. 2023. [Taqyim: Evaluating arabic nlp tasks using chatgpt models](https://md.research.elamur.ai/free/en/papers/alyafeai-2023-taqyim-evaluating-arabic-nlp-tasks/).
6. Peiyi Wang, Xiao Bi, et al. 2025. [Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning](https://md.research.elamur.ai/free/en/papers/deepseek-ai-2025-deepseek-r1-incentivizing-reasoning-capability/). arXiv preprint arXiv:2501.12948. [↗](https://arxiv.org/abs/2501.12948)
7. Rohan Anil, Andrew M. Dai, Orhan Firat, Melvin Johnson, Dmitry Lepikhin, Alexandre Passos, Siamak Shakeri, Emanuel Taropa, Paige Bailey, Zhifeng Chen, Eric Chu, Jonathan H. Clark, Laurent El Shafey, Yanping Huang, Kathy Meier-Hellstern, Gaurav Mishra, Erica Moreira, Mark Omernick, Kevin Robinson, Sebastian Ruder, Yi Tay, Kefan Xiao, Yuanzhong Xu, Yujing Zhang, Gustavo Hernandez Abrego, Junwhan Ahn, Jacob Austin, Paul Barham, Jan Botha, James Bradbury, Siddhartha Brahma, Kevin Brooks, Michele Catasta, Yong Cheng, Colin Cherry, Christopher A. Choquette-Choo, Aakanksha Chowdhery, Clément Crepy, Shachi Dave, Mostafa Dehghani, Sunipa Dev, Jacob Devlin, Mark Dı́az, Nan Du, Ethan Dyer, Vlad Feinberg, Fangxiaoyu Feng, Vlad Fienber, Markus Freitag, Xavier Garcia, Sebastian Gehrmann, Lucas Gonzalez, Guy Gur-Ari, Steven Hand, Hadi Hashemi, Le Hou, Joshua Howland, Andrea Hu, Jeffrey Hui, Jeremy Hurwitz, Michael Isard, Abe Ittycheriah, Matthew Jagielski, Wenhao Jia, Kathleen Kenealy, Maxim Krikun, Sneha Kudugunta, Chang Lan, Katherine Lee, Benjamin Lee, Eric Li, Music Li, Wei Li, YaGuang Li, Jian Li, Hyeontaek Lim, Hanzhao Lin, Zhongtao Liu, Frederick Liu, Marcello Maggioni, Aroma Mahendru, Joshua Maynez, Vedant Misra, Maysam Moussalem, Zachary Nado, John Nham, Eric Ni, Andrew Nystrom, Alicia Parrish, Marie Pellat, Martin Polacek, Alex Polozov, Reiner Pope, Siyuan Qiao, Emily Reif, Bryan Richter, Parker Riley, Alex Castro Ros, Aurko Roy, Brennan Saeta, Rajkumar Samuel, Renee Shelby, Ambrose Slone, Daniel Smilkov, David R. So, Daniel Sohn, Simon Tokumine, Dasha Valter, Vijay Vasudevan, Kiran Vodrahalli, Xuezhi Wang, Pidong Wang, Zirui Wang, Tao Wang, John Wieting, Yuhuai Wu, Kelvin Xu, Yunhan Xu, Linting Xue, Pengcheng Yin, Jiahui Yu, Qiao Zhang, Steven Zheng, Ce Zheng, Weikang Zhou, Denny Zhou, Slav Petrov, and Yonghui Wu. 2023. [Palm 2 technical report](https://md.research.elamur.ai/free/en/papers/anil-2023-palm-2-technical-report/).
8. Pengcheng He, Xiaodong Liu, Jianfeng Gao, and Weizhu Chen. 2020. [Deberta: Decoding-enhanced bert with disentangled attention](https://md.research.elamur.ai/free/en/papers/he-2020-deberta-decoding-enhanced-bert-disentangled/). arXiv preprint arXiv:2006.03654. [↗](https://arxiv.org/abs/2006.03654)
9. Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, Weizhu Chen, et al. 2022. [Lora: Low-rank adaptation of large language models](https://md.research.elamur.ai/free/en/papers/hu-2021-lora-low-rank-adaptation-large/). ICLR, 1(2):3.
10. Shiyuan Huang, Siddarth Mamidanna, Shreedhar Jangam, Yilun Zhou, and Leilani H Gilpin. 2023. Can large language models explain themselves? a study of llm-generated self-explanations. arXiv preprint arXiv:2310.11207. [↗](https://arxiv.org/abs/2310.11207)
11. Aaron Hurst, Adam Lerer, Adam P Goucher, Adam Perelman, Aditya Ramesh, Aidan Clark, AJ Ostrow, Akila Welihinda, Alan Hayes, Alec Radford, et al. 2024. [Gpt-4o system card](https://md.research.elamur.ai/free/en/papers/openai-2024-gpt-4o-system-card/). arXiv preprint arXiv:2410.21276. [↗](https://arxiv.org/abs/2410.21276)
12. Md Tawkat Islam Khondaker, Abdul Waheed, El Moatez Billah Nagoudi, and Muhammad Abdul-Mageed. 2023. [Gptaraeval: A comprehensive evaluation of chatgpt on arabic nlp](https://md.research.elamur.ai/free/en/papers/khondaker-2023-gptaraeval-comprehensive-evaluation-chatgpt-arabic/). arXiv preprint arXiv:2305.14976. [↗](https://arxiv.org/abs/2305.14976)
13. Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019. [Roberta: A robustly optimized bert pretraining approach](https://md.research.elamur.ai/free/en/papers/liu-2019-roberta-robustly-optimized-bert-pretraining/). arXiv preprint arXiv:1907.11692. [↗](https://arxiv.org/abs/1907.11692)
14. Jenish Maharjan, Anurag Garikipati, Navan Preet Singh, Leo Cyrus, Mayank Sharma, Madalina Ciobanu, Gina Barnes, Rahul Thapa, Qingqing Mao, and Ritankar Das. 2024. Openmedlm: prompt engineering can out-perform fine-tuning in medical question-answering with open-source large language models. Scientific Reports, 14(1):14156. [find](https://scholar.google.com/scholar?q=Jenish+Maharjan%2C+Anurag+Garikipati%2C+Navan+Preet+Singh%2C+Leo+Cyrus%2C+Mayank+Sharma%2C+Madalina+Ciobanu%2C+Gina+Barnes%2C+Rahul+Thapa%2C+Qingqing+Mao%2C+and+Ritankar+Das%2E+2024%2E+Openmedlm%3A+prompt+engineering+can+out)
15. Wissam Antoun, Fady Baly, and Hazem Hajj. 2020. [Arabert: Transformer-based model for arabic language understanding](https://md.research.elamur.ai/free/en/papers/antoun-2020-arabert-transformer-based-model-arabic/). In Proc. OSACT, pages 9–15.
16. Qwen, :, An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, Huan Lin, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxi Yang, Jingren Zhou, Junyang Lin, Kai Dang, Keming Lu, Keqin Bao, Kexin Yang, Le Yu, Mei Li, Mingfeng Xue, Pei Zhang, Qin Zhu, Rui Men, Runji Lin, Tianhao Li, Tianyi Tang, Tingyu Xia, Xingzhang Ren, Xuancheng Ren, Yang Fan, Yang Su, Yichang Zhang, Yu Wan, Yuqiong Liu, Zeyu Cui, Zhenru Zhang, and Zihan Qiu. 2025. Qwen2.5 technical report. [find](https://scholar.google.com/scholar?q=Qwen%2C+%3A%2C+An+Yang%2C+Baosong+Yang%2C+Beichen+Zhang%2C+Binyuan+Hui%2C+Bo+Zheng%2C+Bowen+Yu%2C+Chengyuan+Li%2C+Dayiheng+Liu%2C+Fei+Huang%2C+Haoran+Wei%2C+Huan+Lin%2C+Jian+Yang%2C+Jianhong+Tu%2C+Jianwei+Zhang%2C+Jianxin+Yang%2C+Jiaxi+)
17. Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. 2020. [Language models are few-shot learners](https://md.research.elamur.ai/free/en/papers/brown-2020-language-models-few-shot-learners/). Advances in neural information processing systems, 33:1877–1901.
18. Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, et al. 2024. [The llama 3 herd of models](https://md.research.elamur.ai/free/en/papers/grattafiori-2024-llama-3-herd-models/). arXiv preprint arXiv:2407.21783. Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, Dan Bikel, Lukas Blecher, Cristian Canton Ferrer, Moya Chen, Guillem Cucurull, David Esiobu, Jude Fernandes, Jeremy Fu, Wenyin Fu, Brian Fuller, Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, [↗](https://arxiv.org/abs/2407.21783)

*Updated 2 November 2026*