Elamur Research⌕
☰
home / papers / 2026 / looploop at SemEval-2026 Task 3: A Dimensional Aspect-Based Sentiment System with DeBERTa Regression and Qwen3 Instruction Fine-Tuning

looploop at SemEval-2026 Task 3: A Dimensional Aspect-Based Sentiment System with DeBERTa Regression and Qwen3 Instruction Fine-Tuning

Paper · 2026 · Liu Yang, Gang Hu and Jing Li

Summary

SemEval-2026 Task 3 (DimABSA) replaces discrete polarity labels with continuous Valence-Arousal (VA) scores on a 1–9 scale, assigned to individual aspects. The looploop team tackles all three subtasks using only the official English laptop and restaurant data: scoring VA for a given aspect (Task 1), extracting aspect–opinion–VA triplets (Task 2), and extracting aspect–category–opinion–VA quadruplets (Task 3).

Rather than using one model for everything, they split the work. Task 1 uses a DeBERTa-v3-large regressor that averages token states with MeanPooling, feeds them to an MLP head whose sigmoid output is rescaled to the VA range, and trains with Huber loss. Tasks 2 and 3 are treated as text generation with Qwen3-4B-Instruct, fine-tuned via 4-bit QLoRA. The generated output then goes through regex parsing, a span recovery step that maps strings back to exact source text, and clipping of VA values.

The system clearly outperforms the official LLM baselines and lands mid-table on the leaderboard. It shows that combining an encoder regressor with a compact fine-tuned LLM is practical, and that exact-span alignment matters most under strict matching.

Key points

  • Task 1 RMSE VA averages 1.2534 (Laptop 1.3021, Restaurant 1.2048). The baselines score 2.1677 (Kimi-K2 Thinking) and 2.7258 (Qwen-3 14B).
  • Average cF1 is 0.5299 on Task 2 and 0.4171 on Task 3. Restaurant Task 3 reaches 0.5562, versus 0.3746 for the best baseline, while Laptop Task 3 (0.2781) is roughly at baseline level.
  • English-track ranks (Laptop/Restaurant) are 6th/5th on Task 1, 16th/14th on Task 2, and 11th/10th on Task 3.
  • In ablations, removing span recovery cuts Task 2 cF1 from 0.5299 to 0.4305. Swapping MeanPooling for [CLS] or Huber for MSE slightly worsens RMSE (1.2725 and 1.2679, respectively).
  • The main error sources are under-prediction of extreme VA values, confusion between categories such as FOOD#QUALITY and RESTAURANT#GENERAL, and implicit aspects. The code is available at github.com/kiss-the-rain/DimABSA2026.

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

🎓Yunnan UniversityKunming · China17 works · 40 citations

 Fields, topics and concepts

📚 Shelves 6The 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 worksPsychology · 1,552 worksExperimental and Cognitive Psychology · 870 works

🧭 Topics 1Research subjects and languages the work is about.Sentiment & emotion · 972 works

Cite

Cite as Liu Yang, Gang Hu and Jing Li. 2026. looploop at SemEval-2026 Task 3: A Dimensional Aspect-Based Sentiment System with DeBERTa Regression and Qwen3 Instruction Fine-Tuning. 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-1.196

@inproceedings{yang2026looploop,
  title = {{looploop at SemEval-2026 Task 3: A Dimensional Aspect-Based Sentiment System with DeBERTa Regression and Qwen3 Instruction Fine-Tuning}},
  author = {Liu Yang, Gang Hu and Jing Li},
  booktitle = {Proceedings of the 20th International Workshop on Semantic Evaluation (2026)},
  year = {2026},
  doi = {10.18653/v1/2026.semeval-1.196},
  url = {https://doi.org/10.18653/v1/2026.semeval-1.196}
}

References

19 references, 12 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Jinze Bai, Shuai Bai, Yunfei Chu, Zeyu Cui, Kai Dang, Xiaodong Deng, Yang Fan, Wenbin Ge, Yu Han, Fei Huang, Binyuan Hui, and Luo Ji. 2023. Qwen technical report. Preprint, arXiv:2309.16609.
  2. Jonas Becker, Liang-Chih Yu, Shamsuddeen Hassan Muhammad, Jan Philip Wahle, Terry Ruas, Lung-Hao Lee, and Saif M. Mohammad. 2026. Dimstance: Multilingual datasets for dimensional stance analysis. Preprint, arXiv:2601.21483. find
  3. Sven Buechel and Udo Hahn. 2017. Emobank: Studying the impact of annotation perspective and representation format on dimensional emotion analysis. In Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 2, Short Papers, pages 578–585. find
  4. Hongjie Cai, Rui Xia, and Jianfei Yu. 2021. Aspect-category-opinion-sentiment quadruple extraction with implicit aspects and opinions. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics (ACL), pages 340–350.
  5. Tim Dettmers, Artidoro Pagnoni, Ari Holtzman, and Luke Zettlemoyer. 2023. Qlora: Efficient finetuning of quantized llms. In Advances in Neural Information Processing Systems.
  6. Pengcheng He, Xiaodong Liu, Jianfeng Gao, and Weizhu Chen. 2021. Deberta: Decoding-enhanced bert with disentangled attention. In International Conference on Learning Representations (ICLR).
  7. Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, and Weizhu Chen. 2022. Lora: Low-rank adaptation of large language models. In International Conference on Learning Representations (ICLR).
  8. Peter J. Huber. 1964. Robust estimation of a location parameter. The Annals of Mathematical Statistics, 35(1):73–101. find
  9. Lung-Hao Lee, Liang-Chih Yu, Natalia Loukashe-vich, Ilseyar Alimova, Alexander Panchenko, Tzu-Mi Lin, Zhe-Yu Xu, Jian-Yu Zhou, Guangmin Zheng, Jin Wang, Sharanya Awasthi, Jonas Becker, Jan Philip Wahle, Terry Ruas, Shamsuddeen Hassan Muhammad, and Saif M. Mohammad. 2026. Dimabsa: Building multilingu… find
  10. Haiyun Peng, Lu Xu, Lidong Bing, Fei Huang, Wei Lu, and Luo Si. 2020. Knowing what, how and why: A near complete solution for aspect-based sentiment analysis. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 8600–8613. find
  11. Maria Pontiki, Dimitrios Galanis, John Pavlopoulos, Harris Papageorgiou, Ion Androutsopoulos, and Suresh Manandhar. 2014. Semeval-2014 task 4: Aspect based sentiment analysis. In Proceedings of the 8th international workshop on semantic evaluation (SemEval 2014), pages 27–35.
  12. Maria Pontiki, Dimitris Galanis, Haris Papageorgiou, Ion Androutsopoulos, Suresh Manandhar, Mohammad AL-Smadi, Mahmoud Al-Ayyoub, Yanyan Zhao, Bing Qin, Orphée De Clercq, Véronique Hoste, Marianna Apidianaki, Xavier Tannier, Natalia Loukachevitch, Evgeniy Kotelnikov, Nuria Bel, Salud María Jiménez-Z… — SemEval-2016 Task 5: Aspect Based Sentiment Analysis
  13. Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. 2020. Exploring the limits of transfer learning with a unified text-to-text transformer. Journal of Machine Learning Research, 21(140):1–67.
  14. Nils Reimers and Iryna Gurevych. 2019. Sentence-bert: Sentence embeddings using siamese bert-networks. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing (EMNLP-IJCNLP).
  15. James A Russell. 1980. A circumplex model of affect. Journal of personality and social psychology, 39(6):1161. find
  16. Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. In Advances in Neural Information Processing Systems (NeurIPS).
  17. Liang-Chih Yu, Jonas Becker, Shamsuddeen Hassan Muhammad, Idris Abdulmumin, Lung-Hao Lee, Ying-Lung Lin, Jin Wang, Jan Philip Wahle, Terry Ruas, Alexander Panchenko, Ilseyar Alimova, Kai-Wei Chang, Lilian Wanzare, Nelson Odhiambo, Bela Gipp, and Saif M. Mohammad. 2026. SemEval-2026 task 3: Dimension… find
  18. Wenxuan Zhang, Yang Deng, Xin Li, Yifei Yuan, Lidong Bing, and Wai Lam. 2021. Aspect sentiment quad prediction as paraphrase generation. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 9209– 9219.
  19. Wenxuan Zhang, Yue Deng, Bing Liu, Sinno Jialin Pan, and Lidong Bing. 2023. Sentiment analysis in the era of large language models: A reality check. arXiv preprint arXiv:2305.15005.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/yang-2026-looploop-semeval-2026-task-3/