---
title: "CSE-UOI at SemEval-2026 Task 6: A Two-Stage Heterogeneous"
description: "This paper addresses Subtask 1 of SemEval-2026 Task 6. 0.8505 accuracy. 16 references."
authors: ["Christos Tzouvaras", "Konstantinos Skianis", "Athanasios Voulodimos"]
year: 2026
venue: "Proceedings of the 20th International Workshop on Semantic Evaluation (2026)"
source: "https://aclanthology.org/2026.semeval-1.435/"
doi: "10.18653/v1/2026.semeval-1.435"
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/tzouvaras-2026-cse-uoi-semeval-2026-task/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/tzouvaras-2026-cse-uoi-semeval-2026-task/)

[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/26/) / CSE-UOI at SemEval-2026 Task 6: A Two-Stage Heterogeneous Ensemble with Deliberative Complexity Gating for Political Evasion Detection

# CSE-UOI at SemEval-2026 Task 6: A Two-Stage Heterogeneous Ensemble with Deliberative Complexity Gating for Political Evasion Detection

**Authors:** Christos Tzouvaras, [Konstantinos Skianis](https://md.research.elamur.ai/free/en/authors/konstantinos-skianis/) and [Athanasios Voulodimos](https://md.research.elamur.ai/free/en/authors/athanasios-voulodimos/) · **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-1.435/) · [PDF](https://aclanthology.org/2026.semeval-1.435.pdf) · [DOI 10.18653/v1/2026.semeval-1.435](https://doi.org/10.18653/v1/2026.semeval-1.435) · [OpenAlex](https://openalex.org/W7166856133)

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

## Summary
This paper addresses Subtask 1 of SemEval-2026 Task 6. The goal is to decide whether a politician's answer in an English interview question–answer pair is a Clear Reply, Ambivalent, or Clear Non-Reply. The authors use an inference-only pipeline with no fine-tuning, built in two stages. In Stage 1, two different commercial LLMs, grok-4-1-fast-reasoning and gemini-3-flash-preview, each produce five Chain-of-Thought samples over the nine fine-grained evasion labels. These labels are mapped onto the three clarity classes and merged through asymmetric weighted voting that favours Grok. Stage 2, Deliberative Complexity Gating (DCG), is an unsupervised post-hoc rule that needs no additional API calls. When Gemini's average output exceeds the batch's first-quartile length and Grok's samples disagree, Gemini's vote is changed to Ambivalent and the vote is recomputed. DCG raises evaluation Macro-F1 from 0.8122 to 0.8505, which placed 3rd of 41 submissions. The work suggests that simple behavioural cues such as output length can cheaply identify ambiguous cases. Open-weight replications indicate the effect does not depend on proprietary models.

## Key points
- Data: the official QEvasion resources, with a labelled development split of 308 pairs and a blind evaluation split of 237 pairs. Code is released on GitHub.
- Evaluation Macro-F1 by configuration:

  | Configuration | Macro-F1 |
  |---|---|
  | Grok-only | 0.8264 |
  | Gemini-only | 0.8083 |
  | Stage 1 | 0.8122 |
  | Stage 2 (DCG) | 0.8505 |

  Accuracy rose from 81.0% to 84.4%. The final score tied with AsymVerify for the second-best result, behind only TeleAI at 0.89.
- Predicting evasion labels first acts as an error buffer: 58.5% of evasion-level mistakes still land in the correct clarity class.
- Error analysis: 86.8% of the mistakes that remain involve confusion between the Clear Reply and Ambivalent classes. About 10.1% of development samples are misclassified by every system variant.
- Open-source transfer:
  - With Gemma 4 in Grok's role and Kimi K2.6 in Gemini's role, evaluation Macro-F1 improved from 0.8317 to 0.8364.
  - Pairing Grok with Kimi, DeepSeek, or Qwen 3.6 also produced DCG gains, though smaller ones.

*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 Technical University of Athens](https://md.research.elamur.ai/free/en/universities/national-technical-university-of-athens/) — University, Athens, Greece · 8 works · 461 citations
- [University of Ioannina](https://md.research.elamur.ai/free/en/universities/university-of-ioannina/) — University, Ioannina, Greece · 2 works · 2 citations

📚 Shelves (6): [🖥️ 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 · [🤝 Social Sciences](https://md.research.elamur.ai/free/en/shelves/social-sciences/) · 1,772 works · [🤝 General Social Sciences](https://md.research.elamur.ai/free/en/shelves/social-sciences/general-social-sciences/) · 227 works

## Cite

Cite as: Christos Tzouvaras, Konstantinos Skianis and Athanasios Voulodimos. 2026. CSE-UOI at SemEval-2026 Task 6: A Two-Stage Heterogeneous Ensemble with Deliberative Complexity Gating for Political Evasion 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.

```bibtex
@inproceedings{tzouvaras2026cse,
  title = {{CSE-UOI at SemEval-2026 Task 6: A Two-Stage Heterogeneous Ensemble with Deliberative Complexity Gating for Political Evasion Detection}},
  author = {Christos Tzouvaras, Konstantinos Skianis and Athanasios Voulodimos},
  booktitle = {Proceedings of the 20th International Workshop on Semantic Evaluation (2026)},
  year = {2026},
  doi = {10.18653/v1/2026.semeval-1.435},
  url = {https://doi.org/10.18653/v1/2026.semeval-1.435}
}
```

- [← previous CS-YODAS: A Mined Dataset of In-the-Wild Code-Switched Speech](https://md.research.elamur.ai/free/en/papers/yan-2026-cs-yodas-mined-dataset-wild/)
- [next → CSECU-DSG at SemEval-2026 Task 6: Imbalance-Aware Transformers for Unmasking Political Question Evasions](https://md.research.elamur.ai/free/en/papers/shesgin-2026-csecu-dsg-semeval-2026-task/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/26/)

## References

16 references, 5 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Sangmin Bae, Jongwoo Ko, Hwanjun Song, and Se-Young Yun. 2023. Fast and robust early-exiting framework for autoregressive language models with synchronized parallel decoding. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 5910–5924, Singapore. Associ… [find](https://scholar.google.com/scholar?q=Sangmin+Bae%2C+Jongwoo+Ko%2C+Hwanjun+Song%2C+and+Se%2DYoung+Yun%2E+2023%2E+Fast+and+robust+early%2Dexiting+framework+for+autoregressive+language+models+with+synchronized+parallel+decoding%2E+In+Proceedings+of+the+202)
2. Peter Bull. 2003. The microanalysis of political communication: Claptrap and ambiguity. Routledge. [find](https://scholar.google.com/scholar?q=Peter+Bull%2E+2003%2E+The+microanalysis+of+political+communication%3A+Claptrap+and+ambiguity%2E+Routledge%2E)
3. Zhichen Dong, Zhanhui Zhou, Zhixuan Liu, Chao Yang, and Chaochao Lu. 2025. Emergent response planning in LLMs. In Forty-second International Conference on Machine Learning. [find](https://scholar.google.com/scholar?q=Zhichen+Dong%2C+Zhanhui+Zhou%2C+Zhixuan+Liu%2C+Chao+Yang%2C+and+Chaochao+Lu%2E+2025%2E+Emergent+response+planning+in+LLMs%2E+In+Forty%2Dsecond+International+Conference+on+Machine+Learning%2E)
4. Yilun Du, Shuang Li, Antonio Torralba, Joshua B. Tenenbaum, and Igor Mordatch. 2024. [Improving factuality and reasoning in language models through multiagent debate](https://md.research.elamur.ai/free/en/papers/du-2023-improving-factuality-reasoning-language-models/). In Proceedings of the 41st International Conference on Machine Learning, volume 235 of Proceedings of Machine Learning Research, pages…
5. Baraa Hikal, Mohamed Basem, Islam Oshallah, and Ali Hamdi. 2025. MSA at BEA 2025 shared task: Disagreement-aware instruction tuning for multidimensional evaluation of LLMs as math tutors. In Proceedings of the 20th Workshop on Innovative Use of NLP for Building Educational Applications (BEA 2025), p… [find](https://scholar.google.com/scholar?q=Baraa+Hikal%2C+Mohamed+Basem%2C+Islam+Oshallah%2C+and+Ali+Hamdi%2E+2025%2E+MSA+at+BEA+2025+shared+task%3A+Disagreement%2Daware+instruction+tuning+for+multidimensional+evaluation+of+LLMs+as+math+tutors%2E+In+Proceedin)
6. Tianyu Hu, Zhen Tan, Song Wang, Huaizhi Qu, and Tianlong Chen. 2025. Multi-agent debate for LLM judges with adaptive stability detection. In The Thirty-ninth Annual Conference on Neural Information Processing Systems. [find](https://scholar.google.com/scholar?q=Tianyu+Hu%2C+Zhen+Tan%2C+Song+Wang%2C+Huaizhi+Qu%2C+and+Tianlong+Chen%2E+2025%2E+Multi%2Dagent+debate+for+LLM+judges+with+adaptive+stability+detection%2E+In+The+Thirty%2Dninth+Annual+Conference+on+Neural+Information+Pr)
7. Denis Janiak, Jakub Binkowski, Albert Sawczyn, Bogdan Gabrys, Ravid Shwartz-Ziv, and Tomasz Jan Kajdanowicz. 2025. The illusion of progress: Re-evaluating hallucination detection in LLMs. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 34728– 34745, S… [find](https://scholar.google.com/scholar?q=Denis+Janiak%2C+Jakub+Binkowski%2C+Albert+Sawczyn%2C+Bogdan+Gabrys%2C+Ravid+Shwartz%2DZiv%2C+and+Tomasz+Jan+Kajdanowicz%2E+2025%2E+The+illusion+of+progress%3A+Re%2Devaluating+hallucination+detection+in+LLMs%2E+In+Proceedin)
8. Tian Liang, Zhiwei He, Wenxiang Jiao, Xing Wang, Yan Wang, Rui Wang, Yujiu Yang, Shuming Shi, and Zhaopeng Tu. 2024. [Encouraging divergent thinking in large language models through multi-agent debate](https://md.research.elamur.ai/free/en/papers/liang-2023-encouraging-divergent-thinking-large-language/). In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 17…
9. Jie Ren, Yao Zhao, Tu Vu, Peter J. Liu, and Balaji Lakshminarayanan. 2023. Self-evaluation improves selective generation in large language models. In Proceedings on "I Can’t Believe It’s Not Better: Failure Modes in the Age of Foundation Models" at NeurIPS 2023 Workshops, volume 239 of Proceedings o… [find](https://scholar.google.com/scholar?q=Jie+Ren%2C+Yao+Zhao%2C+Tu+Vu%2C+Peter+J%2E+Liu%2C+and+Balaji+Lakshminarayanan%2E+2023%2E+Self%2Devaluation+improves+selective+generation+in+large+language+models%2E+In+Proceedings+on+%22I+Can%E2%80%99t+Believe+It%E2%80%99s+Not+Better%3A+F)
10. Tal Schuster, Adam Fisch, Jai Gupta, Mostafa Dehghani, Dara Bahri, Vinh Tran, Yi Tay, and Donald Metzler. 2022. Confident adaptive language modeling. In Advances in Neural Information Processing Systems, volume 35, pages 17456–17472. Curran Associates, Inc. [find](https://scholar.google.com/scholar?q=Tal+Schuster%2C+Adam+Fisch%2C+Jai+Gupta%2C+Mostafa+Dehghani%2C+Dara+Bahri%2C+Vinh+Tran%2C+Yi+Tay%2C+and+Donald+Metzler%2E+2022%2E+Confident+adaptive+language+modeling%2E+In+Advances+in+Neural+Information+Processing+Syste)
11. Konstantinos Thomas, Giorgos Filandrianos, Maria Lymperaiou, Chrysoula Zerva, and Giorgos Stamou. 2024. “I never said that”: A dataset, taxonomy and baselines on response clarity classification. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 5204–5233, Miami, Florida… [find](https://scholar.google.com/scholar?q=Konstantinos+Thomas%2C+Giorgos+Filandrianos%2C+Maria+Lymperaiou%2C+Chrysoula+Zerva%2C+and+Giorgos+Stamou%2E+2024%2E+%E2%80%9CI+never+said+that%E2%80%9D%3A+A+dataset%2C+taxonomy+and+baselines+on+response+clarity+classification%2E+In+Fi)
12. Konstantinos Thomas, Giorgos Filandrianos, Maria Lymperaiou, Chrysoula Zerva, and Giorgos Stamou. 2026. [Semeval-2026 task 6: Clarity – unmasking political question evasions](https://md.research.elamur.ai/free/en/papers/thomas-2026-semeval-2026-task-6-clarity/). Preprint, arXiv:2603.14027.
13. Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc V Le, Ed H. Chi, Sharan Narang, Aakanksha Chowdhery, and Denny Zhou. 2023. [Self-consistency improves chain of thought reasoning in language models](https://md.research.elamur.ai/free/en/papers/wang-2022-self-consistency-improves-chain-thought/). In The Eleventh International Conference on Learning Representations.
14. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, brian ichter, Fei Xia, Ed Chi, Quoc V Le, and Denny Zhou. 2022. [Chain-of-thought prompting elicits reasoning in large language models](https://md.research.elamur.ai/free/en/papers/wei-2022-chain-thought-prompting-elicits-reasoning/). In Advances in Neural Information Processing Systems, volume 35, pages 24824–24837. Curran Associates, Inc.
15. Jiaxin Zhang. 2025. Confidence-aware reasoning: Optimizing self-guided thinking trajectories in large reasoning models. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing: Industry Track, pages 2081–2095, Suzhou (China). Association for Computational Linguistic… [find](https://scholar.google.com/scholar?q=Jiaxin+Zhang%2E+2025%2E+Confidence%2Daware+reasoning%3A+Optimizing+self%2Dguided+thinking+trajectories+in+large+reasoning+models%2E+In+Proceedings+of+the+2025+Conference+on+Empirical+Methods+in+Natural+Language+P)
16. James Xu Zhao, Jimmy Z.j. Liu, Bryan Hooi, and See-Kiong Ng. 2025. How does response length affect long-form factuality. In Findings of the Association for Computational Linguistics: ACL 2025, pages 3102–3125, Vienna, Austria. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=James+Xu+Zhao%2C+Jimmy+Z%2Ej%2E+Liu%2C+Bryan+Hooi%2C+and+See%2DKiong+Ng%2E+2025%2E+How+does+response+length+affect+long%2Dform+factuality%2E+In+Findings+of+the+Association+for+Computational+Linguistics%3A+ACL+2025%2C+pages+3)

*Updated 1 November 2026*