---
title: "kevinyu66 at SemEval-2026 Task 3 — paper by Yu and Liu (2026)"
description: "The authors describe the examples as \"subjective anchors\" that help calibrate VA values to the… 18 references, cited by 1 paper here."
authors: ["Kuanlin Yu", "Wen-Ni Liu"]
year: 2026
venue: "Proceedings of the 20th International Workshop on Semantic Evaluation (2026)"
source: "https://aclanthology.org/2026.semeval-1.16/"
doi: "10.18653/v1/2026.semeval-1.16"
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/yu-2026-kevinyu66-semeval-2026-task-3/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/yu-2026-kevinyu66-semeval-2026-task-3/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/14/) / kevinyu66 at SemEval-2026 Task 3: A Retrieval-Augmented LLM System for Aspect–Opinion Triplet Extraction

# kevinyu66 at SemEval-2026 Task 3: A Retrieval-Augmented LLM System for Aspect–Opinion Triplet Extraction

**Authors:** Kuanlin Yu and Wen-Ni Liu · **Year:** 2026 · **Venue:** [Proceedings of the 20th International Workshop on Semantic Evaluation (2026)](https://md.research.elamur.ai/free/en/venues/semeval/) ([SemEval 2026](https://md.research.elamur.ai/free/en/venues/semeval/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.semeval-1.16/) · [PDF](https://aclanthology.org/2026.semeval-1.16.pdf) · [DOI 10.18653/v1/2026.semeval-1.16](https://doi.org/10.18653/v1/2026.semeval-1.16) · [OpenAlex](https://openalex.org/W7166812225)

**Citations:** 0 citations (OpenAlex, 2026-10-14) · 1 in our library, of the 149,439 papers whose bibliographies we read

## Summary
The paper addresses Subtask 2 (DimASTE) of SemEval-2026 Task 3 on Dimensional ABSA. In this subtask, a system must extract aspect term, opinion term and a real-valued Valence–Arousal (VA) score from reviews in several languages and domains. The authors identify two difficulties: sentiment labelling is subjective, and a single fine-tuned model adapts poorly to varied unseen inputs.

Their system combines supervised fine-tuning with in-context learning. Qwen2.5-14B-Instruct is adapted with 4-bit LoRA to output triplets as strict JSON. At inference, the system embeds each query with paraphrase-multilingual-mpnet-base-v2 and retrieves the most similar training sentences by cosine similarity. These sentences and their gold triplets are added to the prompt as demonstrations. The authors describe the examples as "subjective anchors" that help calibrate VA values to the dataset's labelling style.

The system beats the official baseline in every evaluated language–domain pair and lands mid-table on the leaderboards. The gains are notable for lower-resource languages, including Ukrainian and Tatar. The work shows that cheap retrieval can improve a fine-tuned mid-sized LLM on continuous sentiment extraction.

## Key points
- **Training and inference setup:**
  - Backbone is Qwen2.5-14B-Instruct, fine-tuned with LoRA on attention projections (r=16, α=32, NF4 quantization) for 3 epochs with lr 2e-4.
  - Inference uses greedy decoding and top-K=10 retrieval on a single A100 40GB.
  - Code is available at github.com/Kevinyu66/dimaste.
- **Evaluation metric:** continuous F1 (cF1). A prediction counts as a true positive only if the aspect and opinion match exactly. Its credit is then reduced by the normalized Euclidean distance between predicted and gold VA.
- **Official test cF1 and ranks:**

| Language | Domain | cF1 | Rank |
|---|---|---|---|
| English | Restaurant | 0.6707 | 5/20 |
| English | Laptop | 0.5503 | 9/20 |
| Japanese | Hotel | 0.5366 | 6/17 |
| Russian | Restaurant | 0.5117 | 9/17 |
| Chinese | Restaurant | 0.5089 | 7/15 |
| Chinese | Laptop | 0.4802 | 6/15 |
| Ukrainian | Restaurant | 0.4865 | 9/15 |
| Tatar | Restaurant | 0.3731 | 10/16 |

- **Ablation on the Chinese laptop dev set:** cF1 rises from 0.4832 (K=0) to 0.4982 (K=3) and 0.5009 (K=10), with diminishing returns.
- **Error analysis:**
  - The most frequent errors are VA miscalibration (22.1%) and opinion span mismatches (20.5%). The authors attribute many span errors to the strict exact-match evaluation.
  - After fine-tuning, the model leans toward recall and sometimes produces extra triplets (7.3%).

*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

- [National Cheng Kung University](https://md.research.elamur.ai/free/en/universities/national-cheng-kung-university/) — University, Tainan, Taiwan · 5 works · 615 citations
- [National Yang Ming Chiao Tung University](https://md.research.elamur.ai/free/en/universities/national-yang-ming-chiao-tung-university/) — University, Hsinchu, Taiwan · 5 works · 85 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works

🧭 Topics (2): [🧠 Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/) · 3,320 works · [🔍 Information retrieval & search](https://md.research.elamur.ai/free/en/topics/information-retrieval/) · 473 works

## Cite

Cite as: Kuanlin Yu and Wen-Ni Liu. 2026. kevinyu66 at SemEval-2026 Task 3: A Retrieval-Augmented LLM System for Aspect–Opinion Triplet Extraction. 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.

```bibtex
@inproceedings{yu2026kevinyu66,
  title = {{kevinyu66 at SemEval-2026 Task 3: A Retrieval-Augmented LLM System for Aspect–Opinion Triplet Extraction}},
  author = {Kuanlin Yu and Wen-Ni Liu},
  booktitle = {Proceedings of the 20th International Workshop on Semantic Evaluation (2026)},
  year = {2026},
  doi = {10.18653/v1/2026.semeval-1.16},
  url = {https://doi.org/10.18653/v1/2026.semeval-1.16}
}
```

- [← previous KDW at SemEval-2026 Task 12: Logic-Driven Distillation with Knowledge Graphs for Efficient Abductive Reasoning](https://md.research.elamur.ai/free/en/papers/zhu-2026-kdw-semeval-2026-task-12/)
- [next → KG-MuLQA: A Framework for KG-based Multi-Level QA Extraction and Long-Context LLM Evaluation](https://md.research.elamur.ai/free/en/papers/tatarinov-2026-kg-mulqa-framework-kg-based/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/14/)

## References

18 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Gianni Brauwers and Flavius Frasincar. 2022. A survey on aspect-based sentiment classification. ACM Comput. Surv., 55(4). [find](https://scholar.google.com/scholar?q=Gianni+Brauwers+and+Flavius+Frasincar%2E+2022%2E+A+survey+on+aspect%2Dbased+sentiment+classification%2E+ACM+Comput%2E+Surv%2E%2C+55%284%29%2E)
2. Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter,… — [Language Models are Few-Shot Learners](https://md.research.elamur.ai/free/en/papers/brown-2020-language-models-few-shot-learners/)
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, pages 578–585. [find](https://scholar.google.com/scholar?q=Sven+Buechel+and+Udo+Hahn%2E+2017%2E+Emobank%3A+Studying+the+impact+of+annotation+perspective+and+representation+format+on+dimensional+emotion+analysis%2E+In+Proceedings+of+the+15th+Conference+of+the+European)
4. Zhibin Gou, Qingyan Guo, and Yujiu Yang. 2023. MvP: Multi-view prompting improves aspect sentiment tuple prediction. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 4380–4397, Toronto, Canada. Association for Computational Lin… [find](https://scholar.google.com/scholar?q=Zhibin+Gou%2C+Qingyan+Guo%2C+and+Yujiu+Yang%2E+2023%2E+MvP%3A+Multi%2Dview+prompting+improves+aspect+sentiment+tuple+prediction%2E+In+Proceedings+of+the+61st+Annual+Meeting+of+the+Association+for+Computational+Ling)
5. Mengting Hu, Yike Wu, Hang Gao, Yinhao Bai, and Shiwan Zhao. 2022. Improving aspect sentiment quad prediction via template-order data augmentation. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 7889–7900, Abu Dhabi, United Arab Emirates. Association… [find](https://scholar.google.com/scholar?q=Mengting+Hu%2C+Yike+Wu%2C+Hang+Gao%2C+Yinhao+Bai%2C+and+Shiwan+Zhao%2E+2022%2E+Improving+aspect+sentiment+quad+prediction+via+template%2Dorder+data+augmentation%2E+In+Proceedings+of+the+2022+Conference+on+Empirical+M)
6. Zhongquan Jian, Yanhao Chen, Jiajian Li, Shaopan Wang, Xiangjian Zeng, Junfeng Yao, Xinying An, and Qingqiang Wu. 2025. Simrp: syntactic and semantic similarity retrieval prompting enhances aspect sentiment quad prediction. In Proceedings of the Thirty-Ninth AAAI Conference on Artificial Intelligenc… [find](https://scholar.google.com/scholar?q=Zhongquan+Jian%2C+Yanhao+Chen%2C+Jiajian+Li%2C+Shaopan+Wang%2C+Xiangjian+Zeng%2C+Junfeng+Yao%2C+Xinying+An%2C+and+Qingqiang+Wu%2E+2025%2E+Simrp%3A+syntactic+and+semantic+similarity+retrieval+prompting+enhances+aspect+sen)
7. 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](https://scholar.google.com/scholar?q=Lung%2DHao+Lee%2C+Jian%2DHong+Li%2C+and+Liang%2DChih+Yu%2E+2022%2E+Chinese+emobank%3A+Building+valence%2Darousal+resources+for+dimensional+sentiment+analysis%2E+ACM+Transactions+on+Asian+and+Low%2DResource+Language+Informa)
8. 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](https://scholar.google.com/scholar?q=Lung%2DHao+Lee%2C+Liang%2DChih+Yu%2C+Natalia+Loukashe%2Dvich%2C+Ilseyar+Alimova%2C+Alexander+Panchenko%2C+Tzu%2DMi+Lin%2C+Zhe%2DYu+Xu%2C+Jian%2DYu+Zhou%2C+Guangmin+Zheng%2C+Jin+Wang%2C+Sharanya+Awasthi%2C+Jonas+Becker%2C+Jan+Philip+Wahl)
9. Haoyue Liu, Ishani Chatterjee, MengChu Zhou, Xiaoyu Sean Lu, and Abdullah Abusorrah. 2020. Aspect-based sentiment analysis: A survey of deep learning methods. IEEE Transactions on Computational Social Systems, 7(6):1358–1375. [find](https://scholar.google.com/scholar?q=Haoyue+Liu%2C+Ishani+Chatterjee%2C+MengChu+Zhou%2C+Xiaoyu+Sean+Lu%2C+and+Abdullah+Abusorrah%2E+2020%2E+Aspect%2Dbased+sentiment+analysis%3A+A+survey+of+deep+learning+methods%2E+IEEE+Transactions+on+Computational+Social)
10. OpenAI, Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, Red Avila, Igor Babuschkin, Suchir Balaji, Valerie Balcom, Paul Baltescu, Haiming Bao, Mohammad Bavarian, Jeff Belgum, and 262 others.… — [GPT-4 Technical Report](https://md.research.elamur.ai/free/en/papers/openai-2023-gpt-4-technical-report/)
11. Yi Ren and Danica J. Sutherland. 2025. Learning dynamics of LLM finetuning. In The Thirteenth International Conference on Learning Representations. [find](https://scholar.google.com/scholar?q=Yi+Ren+and+Danica+J%2E+Sutherland%2E+2025%2E+Learning+dynamics+of+LLM+finetuning%2E+In+The+Thirteenth+International+Conference+on+Learning+Representations%2E)
12. Zhihao Ruan, Runyang You, Kaifeng Yang, Junxin Lin, Wenwen Dai, Mengyuan Zhou, Meizhi Jin, and Xinyue Mei. 2025. PAI at SemEval-2025 task 11: A large language model ensemble strategy for text-based emotion detection. In Proceedings of the 19th International Workshop on Semantic Evaluation (SemEval-2… [find](https://scholar.google.com/scholar?q=Zhihao+Ruan%2C+Runyang+You%2C+Kaifeng+Yang%2C+Junxin+Lin%2C+Wenwen+Dai%2C+Mengyuan+Zhou%2C+Meizhi+Jin%2C+and+Xinyue+Mei%2E+2025%2E+PAI+at+SemEval%2D2025+task+11%3A+A+large+language+model+ensemble+strategy+for+text%2Dbased+em)
13. James A Russell. 1980. A circumplex model of affect. Journal of personality and social psychology, 39(6):1161. [find](https://scholar.google.com/scholar?q=James+A+Russell%2E+1980%2E+A+circumplex+model+of+affect%2E+Journal+of+personality+and+social+psychology%2C+39%286%29%3A1161%2E)
14. Kevin Scaria, Himanshu Gupta, Siddharth Goyal, Saurabh Sawant, Swaroop Mishra, and Chitta Baral. 2024. InstructABSA: Instruction learning for aspect based sentiment analysis. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human L… [find](https://scholar.google.com/scholar?q=Kevin+Scaria%2C+Himanshu+Gupta%2C+Siddharth+Goyal%2C+Saurabh+Sawant%2C+Swaroop+Mishra%2C+and+Chitta+Baral%2E+2024%2E+InstructABSA%3A+Instruction+learning+for+aspect+based+sentiment+analysis%2E+In+Proceedings+of+the+202)
15. Qwen An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Guanting Dong, Haoran Wei, Huan Lin, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxin Yang, Jingren Zhou, Junyang Lin, and 25 others. 2024. Qwen2.5 technical report. ArXiv, a… [find](https://scholar.google.com/scholar?q=Qwen+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+Guanting+Dong%2C+Haoran+Wei%2C+Huan+Lin%2C+Jian+Yang%2C+Jianhong+Tu%2C+Jianwei+Zhang%2C+Jianxin+Y)
16. 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](https://scholar.google.com/scholar?q=Liang%2DChih+Yu%2C+Jonas+Becker%2C+Shamsuddeen+Hassan+Muhammad%2C+Idris+Abdulmumin%2C+Lung%2DHao+Lee%2C+Ying%2DLung+Lin%2C+Jin+Wang%2C+Jan+Philip+Wahle%2C+Terry+Ruas%2C+Alexander+Panchenko%2C+Ilseyar+Alimova%2C+Kai%2DWei+Chang%2C+Li)
17. Yu Yu, Chao-Han Huck Yang, Jari Kolehmainen, Prashanth G. Shivakumar, Yile Gu, Sungho Ryu Roger Ren, Qi Luo, Aditya Gourav, I-Fan Chen, Yi-Chieh Liu, Tuan Dinh, Ankur Gandhe Denis Filimonov, Shalini Ghosh, Andreas Stolcke, Ariya Ras-tow, and Ivan Bulyko. 2023. Low-rank adaptation of large language m… [find](https://scholar.google.com/scholar?q=Yu+Yu%2C+Chao%2DHan+Huck+Yang%2C+Jari+Kolehmainen%2C+Prashanth+G%2E+Shivakumar%2C+Yile+Gu%2C+Sungho+Ryu+Roger+Ren%2C+Qi+Luo%2C+Aditya+Gourav%2C+I%2DFan+Chen%2C+Yi%2DChieh+Liu%2C+Tuan+Dinh%2C+Ankur+Gandhe+Denis+Filimonov%2C+Shalini+G)
18. Shengyao Zhuang, Bing Liu, Bevan Koopman, and Guido Zuccon. 2023. Open-source large language models are strong zero-shot query likelihood models for document ranking. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 8807–8817, Singapore. Association for Computational L… [find](https://scholar.google.com/scholar?q=Shengyao+Zhuang%2C+Bing+Liu%2C+Bevan+Koopman%2C+and+Guido+Zuccon%2E+2023%2E+Open%2Dsource+large+language+models+are+strong+zero%2Dshot+query+likelihood+models+for+document+ranking%2E+In+Findings+of+the+Association+fo)

*Updated 31 October 2026*