Elamur Research⌕
☰
home / papers / 2026 / MoMo at SemEval-2026 Task 9: Inference-Only Prompting vs. Fine-Tuning for Multilingual Polarization Detection

MoMo at SemEval-2026 Task 9: Inference-Only Prompting vs. Fine-Tuning for Multilingual Polarization Detection

Paper · 2026 · Sushant Ray and Rakshita Saksainaa

Summary

This paper is a SemEval-2026 Task 9 (Subtask 1) system description. The task is to decide whether a text is polarized, across 22 languages, using data adapted from the POLAR dataset. The authors compare three ways of adapting models to the task:

  • Full fine-tuning of multilingual encoders: XLM-RoBERTa Large, which was their official submission, and DeBERTaV3-base.
  • Frozen encoders topped with small trainable residual modules.
  • Inference-only prompting of Qwen LLMs, in zero-shot, random few-shot and retrieval-based few-shot modes.

In the retrieval mode, support examples are chosen by embedding similarity. The authors also track energy use with eco2ai and examine run-to-run variance and per-language results.

Zero-shot prompting falls short of the fine-tuned encoders. Few-shot prompting closes the gap, and similarity-selected examples with Qwen-3-Next-80B slightly beat the fine-tuned baselines. The benefit is largest for lower-resource languages, where encoder accuracy depends heavily on how much labeled data exists. Prompting costs more energy per instance, but the authors argue that skipping task-specific training lowers overall downstream cost. This makes prompting attractive when languages and labels keep shifting over time.

Key points

  • Dataset: POLAR Subtask 1 covers 22 languages, with 73,681 training, 3,687 dev and 33,288 test items. The share of polarized texts varies widely, from 10.7% (hau) to 90.8% (khm).
  • Few-shot retrieval: For each class, the three nearest training examples are pulled using QwenEmbeddings, EmbeddingGemma or LaBSE with cosine similarity. This consistently outperformed random selection.
  • LLMs: The prompted models are Qwen-3-8B and Qwen-3-Next-80B, a mixture-of-experts model with roughly 3B active parameters. Sampling temperature was set to 0.7.
  • Thinking vs. non-thinking: With Qwen-3-Next-80B in retrieval-based few-shot mode, enabling "thinking" raised the 7-language mean macro-F1 from 0.76 to 0.82. Energy use rose roughly tenfold, from 1.98 to 21.05 ×10⁻² Wh.
  • Evaluation setup: Encoders were trained ten times each, and the authors report median macro-F1 with quartiles. Encoders ran on a P100 GPU and the 80B model on an H100. The energy comparison excludes LLM pretraining cost.

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

Also on the byline: Delhi Skill and Entrepreneurship University.

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 works

🧭 Topics 1Research subjects and languages the work is about.Multilingual & cross-lingual NLP · 882 works

Cite

Cite as Sushant Ray and Rakshita Saksainaa. 2026. MoMo at SemEval-2026 Task 9: Inference-Only Prompting vs. Fine-Tuning for Multilingual Polarization Detection. Proceedings of the 20th International Workshop on Semantic Evaluation (2026).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.semeval-1.440

@inproceedings{ray2026momo,
  title = {{MoMo at SemEval-2026 Task 9: Inference-Only Prompting vs. Fine-Tuning for Multilingual Polarization Detection}},
  author = {Sushant Ray and Rakshita Saksainaa},
  booktitle = {Proceedings of the 20th International Workshop on Semantic Evaluation (2026)},
  year = {2026},
  doi = {10.18653/v1/2026.semeval-1.440},
  url = {https://doi.org/10.18653/v1/2026.semeval-1.440}
}

References

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

  1. Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, and 1 others. 2020. Language models are few-shot learners. Advances in neural information processing systems, 33:1877–1901.
  2. Semen Andreevich Budennyy, Vladimir Dmitrievich Lazarev, Nikita Nikolaevich Zakharenko, Aleksei N Korovin, Olga A Plosskaya, Denis Valer’evich Dimitrov, Vladimir S Akhripkin, Ivan V Pavlov, Ivan Valer’evich Oseledets, Ivan Segundovich Barsola, and 1 others. 2022. Eco2ai: carbon emissions tracking of… find
  3. Jack Choquette. 2023. Nvidia hopper h100 gpu: Scaling performance. IEEE Micro, 43(3):9–17. find
  4. Alexis Conneau, Kartikay Khandelwal, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke Zettlemoyer, and Veselin Stoyanov. 2020. Unsupervised cross-lingual representation learning at scale. In Proceedings of the 58th annual meeting of the association fo…
  5. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. Bert: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologies,…
  6. Fangxiaoyu Feng, Yinfei Yang, Daniel Cer, Naveen Arivazhagan, and Wei Wang. 2022. Language-agnostic bert sentence embedding. In Proceedings of the 60th annual meeting of the association for computational linguistics (volume 1: Long papers), pages 878–891.
  7. Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, Amy Yang, Angela Fan, Anirudh Goyal, Anthony Hartshorn, Aobo Yang, Archi Mitra, Archie Sravankumar, Artem Korenev, Arthur Hinsvark, and 542 o… — The Llama 3 Herd of Models
  8. Pengcheng He, Jianfeng Gao, and Weizhu Chen. 2021. Debertav3: Improving deberta using electra-style pre-training with gradient-disentangled embedding sharing. arXiv preprint arXiv:2111.09543.
  9. Neil Houlsby, Andrei Giurgiu, Stanislaw Jastrzebski, Bruna Morrone, Quentin De Laroussilhe, Andrea Gesmundo, Mona Attariyan, and Sylvain Gelly. 2019. Parameter-efficient transfer learning for nlp. In International conference on machine learning, pages 2790–2799. PMLR. find
  10. Junjie Hu, Sebastian Ruder, Aditya Siddhant, Graham Neubig, Orhan Firat, and Melvin Johnson. 2020. Xtreme: A massively multilingual multi-task benchmark for evaluating cross-lingual generalisation. In International conference on machine learning, pages 4411–4421. PMLR. find
  11. Sunbal Iftikhar, Saeed Hamood Alsamhi, and Steven Davy. 2025. Enhancing Sustainability in LLM Training: Leveraging Federated Learning and Parameter-Efficient Fine-Tuning . IEEE Transactions on Sustainable Computing, 10(06):1158–1172. find
  12. Yaobo Liang, Nan Duan, Yeyun Gong, Ning Wu, Fenfei Guo, Weizhen Qi, Ming Gong, Linjun Shou, Daxin Jiang, Guihong Cao, and 1 others. 2020. Xglue: A new benchmark dataset for cross-lingual pre-training, understanding and generation. In Proceedings of the 2020 Conference on Empirical Methods in Natural…
  13. Jiachang Liu, Dinghan Shen, Yizhe Zhang, William B Dolan, Lawrence Carin, and Weizhu Chen. 2022. What makes good in-context examples for gpt-3? In Proceedings of Deep Learning Inside Out (DeeLIO 2022): The 3rd workshop on knowledge extraction and integration for deep learning architectures, pages 10…
  14. Paul Joe Maliakel, Shashikant Ilager, and Ivona Brandic. 2026. Characterizing llm inference energy-performance tradeoffs across workloads and gpu scaling. Preprint, arXiv:2501.08219. find
  15. Sewon Min, Xinxi Lyu, Ari Holtzman, Mikel Artetxe, Mike Lewis, Hannaneh Hajishirzi, and Luke Zettlemoyer. 2022. Rethinking the role of demonstrations: What makes in-context learning work? In Proceedings of the 2022 conference on empirical methods in natural language processing, pages 11048–11064.
  16. Usman Naseem, Robert Geislinger, Juan Ren, Sarah Kohail, Rudy Garrido Veliz, P Sam Sahil, Yiran Zhang, Marco Antonio Stranisci, Idris Abdulmumin, "Ozge Alacam, Cengiz Acar"urk, Aisha Jabr, Saba Anwar, Abinew Ali Ayele, Elena Tutubalina, Aung Kyaw Htet, Xintong Wang, Surendrabikram Thapa, Tanmoy Chak… — SemEval-2026 Task 9: Detecting Multilingual, Multicultural and Multievent Online Polarization
  17. Usman Naseem, Robert Geislinger, Juan Ren, Sarah Kohail, Rudy Garrido Veliz, P Sam Sahil, Yiran Zhang, Marco Antonio Stranisci, Idris Abdulmumin, Özge Alacam, Cengiz Acartürk, Aisha Jabr, Saba Anwar, Abinew Ali Ayele, Simona Frenda, Alessandra Teresa Cignarella, Elena Tutubalina, Oleg Rogov, Aung Ky… — POLAR: A Benchmark for Multilingual, Multicultural, and Multi-Event Online Polarization
  18. NVIDIA. 2016. NVIDIA Tesla P100 GPU Accelerator Datasheet. https://images.nvidia.com/content/tesla/pdf/ nvidia-tesla-p100-PCIe-datasheet.pdf. ↗
  19. Jonas Pfeiffer, Aishwarya Kamath, Andreas Rücklé, Kyunghyun Cho, and Iryna Gurevych. 2021. Adapterfusion: Non-destructive task composition for transfer learning. In Proceedings of the 16th conference of the European chapter of the association for computational linguistics: main volume, pages 487–503…
  20. Alec Radford, Karthik Narasimhan, Tim Salimans, Ilya Sutskever, and 1 others. 2018. Improving language understanding by generative pre-training. find
  21. Ohad Rubin, Jonathan Herzig, and Jonathan Berant. 2022. Learning to retrieve prompts for in-context learning. In Proceedings of the 2022 conference of the North American chapter of the association for computational linguistics: human language technologies, pages 2655–2671.
  22. Sebastian Ruder, Noah Constant, Jan Botha, Aditya Siddhant, Orhan Firat, Jinlan Fu, Pengfei Liu, Junjie Hu, Dan Garrette, Graham Neubig, and 1 others. 2021. Xtreme-r: Towards more challenging and nuanced multilingual evaluation. In Proceedings of the 2021 Conference on Empirical Methods in Natural L…
  23. Roy Schwartz, Jesse Dodge, Noah A Smith, and Oren Etzioni. 2020. Green ai. Communications of the ACM, 63(12):54–63.
  24. Emma Strubell, Ananya Ganesh, and Andrew McCallum. 2019. Energy and policy considerations for deep learning in nlp. In Proceedings of the 57th annual meeting of the association for computational linguistics, pages 3645–3650.
  25. Henrique Schechter Vera, Sahil Dua, Biao Zhang, Daniel Salz, Ryan Mullins, Sindhu Raghuram Panyam, Sara Smoot, Iftekhar Naim, Joe Zou, Feiyang Chen, Daniel Cer, Alice Lisak, Min Choi, Lucas Gonzalez, Omar Sanseviero, Glenn Cameron, Ian Ballantyne, Kat Black, Kaifeng Chen, and 70 others. 2025. Embedd… — [arXiv:2509.20354](https://arxiv.org/abs/2509.20354) find
  26. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, Chujie Zheng, Dayiheng Liu, Fan Zhou, Fei Huang, Feng Hu, Hao Ge, Haoran Wei, Huan Lin, Jialong Tang, and 41 others. 2025. Qwen3 technical report. Preprint, arXiv:2505.09388.
  27. Yanzhao Zhang, Mingxin Li, Dingkun Long, Xin Zhang, Huan Lin, Baosong Yang, Pengjun Xie, An Yang, Dayiheng Liu, Junyang Lin, Fei Huang, and Jingren Zhou. 2025. Qwen3 embedding: Advancing text embedding and reranking through foundation models. Preprint, arXiv:2506.05176.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/ray-2026-momo-semeval-2026-task-9/