Elamur Research⌕
☰
home / papers / 2026 / YangSteam at SemEval-2026 Task 3: Transformer-Based Aspect-Aware Regression for Dimensional Sentiment and Stance Analysis

YangSteam at SemEval-2026 Task 3: Transformer-Based Aspect-Aware Regression for Dimensional Sentiment and Stance Analysis

Paper · 2026 · Tsung-Hsien Yang and Shu-Fei Yang

Summary

The authors describe their system for SemEval-2026 Task 3 (DimABSA). They target Subtask 1 in two tracks: Track A (DimABSA) and Track B (DimStance). The goal is to predict continuous valence–arousal (VA) scores, on a 1–9 scale, for text–aspect pairs in English and Chinese. Track A covers restaurant, laptop and finance reviews. Track B covers stance on environmental protection, with stance targets treated as aspects.

The system uses only standard components:

  • multilingual transformer fine-tuning;
  • [ASP]/[/ASP] markers around the aspect, with the marker token's hidden state replacing [CLS] for pooling;
  • two separate MLP regression heads, one for valence and one for arousal;
  • a 5-fold cross-validation ensemble.

Comparisons on the development set determined the backbone: Track A uses XLM-RoBERTa-large, while Track B relies on mDeBERTa-v3-base.

On every official test set, the system beats the organizer baselines, including LLM prompting with Kimi-K2, Qwen-3 14B and Mistral-3 14B. It does noticeably better on Chinese than on English. The work's value is practical: it shows that backbone choice matters most, and that a smaller encoder can win when training data is limited.

Key points

  • Evaluation metric is RMSE_VA. Track A test scores (unofficial post-evaluation rank in parentheses):
    • E-Res 1.277 (14th)
    • E-Lap 1.346 (11th)
    • C-Res 0.943 (4th)
    • C-Lap 0.687 (8th)
    • C-Fin 0.486 (2nd)
  • Track B test scores: E-Env 1.573 (5th) and C-Env 0.547 (1st). This gives top-5 placement on 4 of 7 subsets.
  • On Track B dev, mDeBERTa-v3-base (86M parameters) beat XLM-R-large (550M), InfoXLM-large and RemBERT, scoring E-Env 2.062 vs 2.291 for XLM-R-large. The authors attribute this to overfitting of larger models on the small stance datasets.
  • Chinese-specific encoders (MacBERT-large, Chinese-RoBERTa-wwm-ext-large, CKIP-BERT) sharply hurt English results. Cross-model ensembles helped only slightly on Chinese, so a single XLM-R-large was submitted for Track A.
  • Training setup:
    • HuggingFace Transformers;
    • learning rates of 2e-5 for the encoder and 1e-4 for the heads;
    • cosine schedule;
    • effective batch size 32;
    • maximum sequence length 256;
    • early stopping with patience 3.

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

🏢Chunghwa TelecomTaipei · Taiwan2 works · 2 citations

 Fields, topics and concepts

📚 Shelves 6The 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 worksPsychology · 1,649 worksExperimental and Cognitive Psychology · 930 works

🧭 Topics 1Research subjects and languages the work is about.Sentiment & emotion · 1,035 works

Cite

Cite as Tsung-Hsien Yang and Shu-Fei Yang. 2026. YangSteam at SemEval-2026 Task 3: Transformer-Based Aspect-Aware Regression for Dimensional Sentiment and Stance Analysis. 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-1.199

@inproceedings{yang2026yangsteam,
  title = {{YangSteam at SemEval-2026 Task 3: Transformer-Based Aspect-Aware Regression for Dimensional Sentiment and Stance Analysis}},
  author = {Tsung-Hsien Yang and Shu-Fei Yang},
  booktitle = {Proceedings of the 20th International Workshop on Semantic Evaluation (2026)},
  year = {2026},
  doi = {10.18653/v1/2026.semeval-1.199},
  url = {https://doi.org/10.18653/v1/2026.semeval-1.199}
}

References

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

  1. Jonas Becker, Liang-Chih Yu, Shamsuddeen Hassan Muhammad, Jan Philip Wahle, Terry Ruas, Idris Abdulmumin, Lung-Hao Lee, Nelson Odhiambo, Lilian Wanzare, Wen-Ni Liu, Tzu-Mi Lin, Zhe-Yu Xu, Ying-Lung Lin, Jin Wang, Maryam Ibrahim Mukhtar, Bela Gipp, and Saif M. Mohammad. 2026. DimStance: Multilingual… find
  2. 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…
  3. Yiming Cui, Wanxiang Che, Ting Liu, Bing Qin, Shijin Wang, and Guoping Hu. 2020. Revisiting pre-trained models for Chinese natural language processing. In Findings of the Association for Computational Linguistics: EMNLP 2020, pages 657–668. Association for Computational Linguistics.
  4. Pengcheng He, Jianfeng Gao, and Weizhu Chen. 2023. DeBERTaV3: Improving DeBERTa using ELECTRA-style pre-training with gradient-disentangled embedding sharing. In Proceedings of the 11th International Conference on Learning Representations.
  5. Pengcheng He, Xiaodong Liu, Jianfeng Gao, and Weizhu Chen. 2021. DeBERTa: Decoding-enhanced BERT with disentangled attention. In Proceedings of the 9th International Conference on Learning Representations.
  6. Lung-Hao Lee, Jian-Hong Li, and Liang-Chih Yu. 2022. Chinese EmoBank: Building valence-arousal resources for dimensional sentiment analysis. ACM Transactions on Asian and Low-Resource Language Information Processing, 21(4):65. find
  7. 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
  8. Ilya Loshchilov and Frank Hutter. 2019. Decoupled weight decay regularization. In Proceedings of the 7th International Conference on Learning Representations. find
  9. Saif M. Mohammad, Felipe Bravo-Marquez, Mohammad Salameh, and Svetlana Kiritchenko. 2018. SemEval-2018 task 1: Affect in tweets. In Proceedings of the 12th International Workshop on Semantic Evaluation (SemEval-2018), pages 1–17. Association for Computational Linguistics.
  10. 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, Evgeny Kotelnikov, Nuria Bel, Salud María Jiménez-Za… — SemEval-2016 Task 5: Aspect Based Sentiment Analysis
  11. James A. Russell. 1980. A circumplex model of affect. Journal of Personality and Social Psychology, 39(6):1161–1178. find
  12. Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, Pier-ric Cistac, Tim Rault, Remi Louf, Morgan Funtowicz, Joe Davison, Sam Shleifer, Patrick von Platen, Clara Ma, Yacine Jernite, Julien Plu, Canwen Xu, Teven Le Scao, Sylvain Gugger, and 3 others. 2020. Transf…
  13. 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
  14. Wenxuan Zhang, Xin Li, Yang Deng, Lidong Bing, and Wai Lam. 2023. A survey on aspect-based sentiment analysis: Tasks, methods, and challenges. IEEE Transactions on Knowledge and Data Engineering, 35(11):11019–11038. find

Updated

↑ back to top

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