---
title: "SG-UniBuc-NLP at SemEval-2026 Task 6: Multi-Head RoBERTa"
description: "The approach offers a memory-bounded, architecture-agnostic alternative to dedicated long-context encoders. 23 references."
authors: ["Gabriel Stefan", "Sergiu Nisioi"]
year: 2026
venue: "Proceedings of the 20th International Workshop on Semantic Evaluation (2026)"
source: "https://aclanthology.org/2026.semeval-1.133/"
doi: "10.18653/v1/2026.semeval-1.133"
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/stefan-2026-sg-unibuc-nlp-semeval-2026/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/stefan-2026-sg-unibuc-nlp-semeval-2026/)

[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/49/) / SG-UniBuc-NLP at SemEval-2026 Task 6: Multi-Head RoBERTa with Chunking for Long-Context Evasion Detection

# SG-UniBuc-NLP at SemEval-2026 Task 6: Multi-Head RoBERTa with Chunking for Long-Context Evasion Detection

**Authors:** Gabriel Stefan and [Sergiu Nisioi](https://md.research.elamur.ai/free/en/authors/sergiu-nisioi/) · **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-1.133/) · [PDF](https://aclanthology.org/2026.semeval-1.133.pdf) · [DOI 10.18653/v1/2026.semeval-1.133](https://doi.org/10.18653/v1/2026.semeval-1.133) · [OpenAlex](https://openalex.org/W7166891709)

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

## Summary
SemEval-2026 Task 6 (CLARITY) targets the automatic detection of evasive answers in English political interviews. Systems label question–answer pairs from White House transcripts with a coarse 3-way Clarity label and a fine 9-way Evasion strategy. About 29% of inputs exceed RoBERTa's 512-token window, so the authors avoid truncation. Each concatenated question–answer sequence is cut into 512-token windows that overlap by half. One shared RoBERTa-large encoder processes every window, and the position-0 vectors are merged through element-wise Max-Pooling. Two linear heads on this pooled vector are trained jointly with a summed cross-entropy loss. Predictions from seven stratified cross-validation models are averaged at inference. The system reached Macro-F1 of 0.80 for Clarity and 0.51 for Evasion, finishing 11th in both subtasks. Max-Pooling outperformed Mean-Pooling and a first-chunk-only baseline, which suggests evasion cues are localized. Errors cluster on rare, pragmatically subtle classes and on items where annotators disagree. Class-weighted and focal losses did not help. The approach offers a memory-bounded, architecture-agnostic alternative to dedicated long-context encoders.

## Key points
- Data: the CLARITY dataset has 3,756 QA pairs from 287 White House interviews (2006–2023), split into 3,448 train and 308 dev instances. Labels are heavily skewed (Ambivalent 59%), and annotator agreement is Fleiss κ 0.64 for Clarity and 0.48 for Evasion.
- Method: windows of L=512 tokens with stride S=256, a shared RoBERTa-large encoder (d=1024), Max-Pooling over chunks, dropout 0.1, two heads with equal-weight losses, and a 7-fold ensemble.
- Results: Macro-F1 of 0.80, ranked 11th of 41 teams (Subtask 1), and 0.51, ranked 11th of 33 teams (Subtask 2).
- Errors: Ambivalent absorbs about 36% of Clear Reply and 31% of Clear Non-Reply cases. Partial/half-answer (79 examples) gets zero out-of-fold accuracy, and Implicit and Deflection are often confused with Explicit or Dodging.
- Future work: data augmentation for minority classes, NLI framing, and ColBERT-style late interaction between question and answer. The code is on GitHub (gabriel-stefan/political-evasion-detection).

*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

- [University of Bucharest](https://md.research.elamur.ai/free/en/universities/university-of-bucharest/) — University, Bucharest, Romania · 21 works · 516 citations

📚 Shelves (6): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works · [🤝 Social Sciences](https://md.research.elamur.ai/free/en/shelves/social-sciences/) · 1,954 works · [🤝 General Social Sciences](https://md.research.elamur.ai/free/en/shelves/social-sciences/general-social-sciences/) · 254 works

## Cite

Cite as: Gabriel Stefan and Sergiu Nisioi. 2026. SG-UniBuc-NLP at SemEval-2026 Task 6: Multi-Head RoBERTa with Chunking for Long-Context 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{stefan2026sg,
  title = {{SG-UniBuc-NLP at SemEval-2026 Task 6: Multi-Head RoBERTa with Chunking for Long-Context Evasion Detection}},
  author = {Gabriel Stefan and Sergiu Nisioi},
  booktitle = {Proceedings of the 20th International Workshop on Semantic Evaluation (2026)},
  year = {2026},
  doi = {10.18653/v1/2026.semeval-1.133},
  url = {https://doi.org/10.18653/v1/2026.semeval-1.133}
}
```

- [← previous Severity-Aware Weighted Loss for Arabic Medical Text Generation](https://md.research.elamur.ai/free/en/papers/alansary-2026-severity-aware-weighted-loss-arabic/)
- [next → SGA-MCTS: Decoupling Planning from Execution via Training-Free Atomic Experience Retrieval](https://md.research.elamur.ai/free/en/papers/xie-2026-sga-mcts-decoupling-planning-execution/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/49/)

## References

23 references, 11 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Iz Beltagy, Matthew E. Peters, and Arman Cohan. 2020. [Longformer: The long-document transformer](https://md.research.elamur.ai/free/en/papers/beltagy-2020-longformer-long-document-transformer/).
2. Peter Bull. 2003. The Microanalysis of Political Communication: Claptrap and Ambiguity. Routledge, London. [find](https://scholar.google.com/scholar?q=Peter+Bull%2E+2003%2E+The+Microanalysis+of+Political+Communication%3A+Claptrap+and+Ambiguity%2E+Routledge%2C+London%2E)
3. Peter Bull and Kate Mayer. 1993. How not to answer questions in political interviews. Political Psychology, 14(4):651–666. [find](https://scholar.google.com/scholar?q=Peter+Bull+and+Kate+Mayer%2E+1993%2E+How+not+to+answer+questions+in+political+interviews%2E+Political+Psychology%2C+14%284%29%3A651%E2%80%93666%2E)
4. Rich Caruana. 1997. Multitask learning. Machine Learning, 28(1):41–75. [find](https://scholar.google.com/scholar?q=Rich+Caruana%2E+1997%2E+Multitask+learning%2E+Machine+Learning%2C+28%281%29%3A41%E2%80%9375%2E)
5. Steven E. Clayman. 2001. Answers and evasions. Language in Society, 30(3):403–442. [find](https://scholar.google.com/scholar?q=Steven+E%2E+Clayman%2E+2001%2E+Answers+and+evasions%2E+Language+in+Society%2C+30%283%29%3A403%E2%80%93442%2E)
6. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. [BERT: Pre-training of deep bidirectional transformers for language understanding](https://md.research.elamur.ai/free/en/papers/devlin-2019-bert-pre-training-deep-bidirectional/). In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies,…
7. Thomas G. Dietterich, Richard H. Lathrop, and Tomás Lozano-Pérez. 1997. Solving the multiple instance problem with axis-parallel rectangles. Artificial Intelligence, 89(1–2):31–71. [find](https://scholar.google.com/scholar?q=Thomas+G%2E+Dietterich%2C+Richard+H%2E+Lathrop%2C+and+Tom%C3%A1s+Lozano%2DP%C3%A9rez%2E+1997%2E+Solving+the+multiple+instance+problem+with+axis%2Dparallel+rectangles%2E+Artificial+Intelligence%2C+89%281%E2%80%932%29%3A31%E2%80%9371%2E)
8. Mikel Galar, Alberto Fernandez, Edurne Barrenechea, Humberto Bustince, and Francisco Herrera. 2012. A review on ensembles for the class imbalance problem: Bagging-, boosting-, and hybrid-based approaches. IEEE Transactions on Systems, Man, and Cybernetics, Part C (Applications and Reviews), 42(4):46… [find](https://scholar.google.com/scholar?q=Mikel+Galar%2C+Alberto+Fernandez%2C+Edurne+Barrenechea%2C+Humberto+Bustince%2C+and+Francisco+Herrera%2E+2012%2E+A+review+on+ensembles+for+the+class+imbalance+problem%3A+Bagging%2D%2C+boosting%2D%2C+and+hybrid%2Dbased+approac)
9. Sandra Harris. 1991. Evasive action: How politicians respond to questions in political interviews. In Paddy Scannell, editor, Broadcast Talk, pages 76–99. Sage, London. [find](https://scholar.google.com/scholar?q=Sandra+Harris%2E+1991%2E+Evasive+action%3A+How+politicians+respond+to+questions+in+political+interviews%2E+In+Paddy+Scannell%2C+editor%2C+Broadcast+Talk%2C+pages+76%E2%80%9399%2E+Sage%2C+London%2E)
10. Maximilian Ilse, Jakub Tomczak, and Max Welling. 2018. Attention-based deep multiple instance learning. In Proceedings of the 35th International Conference on Machine Learning, volume 80, pages 2127– 2136. PMLR. [find](https://scholar.google.com/scholar?q=Maximilian+Ilse%2C+Jakub+Tomczak%2C+and+Max+Welling%2E+2018%2E+Attention%2Dbased+deep+multiple+instance+learning%2E+In+Proceedings+of+the+35th+International+Conference+on+Machine+Learning%2C+volume+80%2C+pages+2127%E2%80%93+)
11. Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019. [Roberta: A robustly optimized bert pretraining approach](https://md.research.elamur.ai/free/en/papers/liu-2019-roberta-robustly-optimized-bert-pretraining/).
12. Ilya Loshchilov and Frank Hutter. 2019. Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101. [find](https://scholar.google.com/scholar?q=Ilya+Loshchilov+and+Frank+Hutter%2E+2019%2E+Decoupled+weight+decay+regularization%2E+arXiv+preprint+arXiv%3A1711%2E05101%2E)
13. Raghavendra Pappagari, Piotr Zelasko, Jesús Villalba, Yishay Carmiel, and Najim Dehak. 2019. Hierarchical transformers for long document classification. In 2019 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU), pages 838–844. IEEE. [find](https://scholar.google.com/scholar?q=Raghavendra+Pappagari%2C+Piotr+Zelasko%2C+Jes%C3%BAs+Villalba%2C+Yishay+Carmiel%2C+and+Najim+Dehak%2E+2019%2E+Hierarchical+transformers+for+long+document+classification%2E+In+2019+IEEE+Automatic+Speech+Recognition+and+U)
14. Parameswary Rasiah. 2010. A framework for the systematic analysis of evasion in parliamentary discourse. Journal of Pragmatics, 42(3):664–680. [find](https://scholar.google.com/scholar?q=Parameswary+Rasiah%2E+2010%2E+A+framework+for+the+systematic+analysis+of+evasion+in+parliamentary+discourse%2E+Journal+of+Pragmatics%2C+42%283%29%3A664%E2%80%93680%2E)
15. Sebastian Ruder. 2017. [An overview of multi-task learning in deep neural networks](https://md.research.elamur.ai/free/en/papers/ruder-2017-overview-multi-task-learning-deep/).
16. Victor Sanh, Lysandre Debut, Julien Chaumond, and Thomas Wolf. 2019. [DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter](https://md.research.elamur.ai/free/en/papers/sanh-2019-distilbert-distilled-version-bert-smaller/). In 5th Workshop on Energy Efficient Machine Learning and Cognitive Computing - NeurIPS 2019.
17. Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov. 2014. [Dropout: A simple way to prevent neural networks from overfitting](https://md.research.elamur.ai/free/en/papers/srivastava-2014-dropout-simple-way-prevent/). Journal of Machine Learning Research, 15(1):1929–1958.
18. 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)
19. 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/).
20. Omar Khattab and Matei Zaharia. 2020. [Colbert: Efficient and effective passage search via contextualized late interaction over bert](https://md.research.elamur.ai/free/en/papers/khattab-2020-colbert-efficient-effective-passage/). In Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 39–48. Association for Computing Machinery.
21. Benjamin Warner, Antoine Chaffin, Benjamin Clavié, Orion Weller, Oskar Hallström, Said Taghadouini, Alexis Gallagher, Raja Biswas, Faisal Ladhak, Tom Aarsen, Griffin Thomas Adams, Jeremy Howard, and Iacopo Poli. 2025. [Smarter, better, faster, longer: A modern bidirectional encoder for fast, memory e…](https://md.research.elamur.ai/free/en/papers/warner-2024-smarter-better-faster-longer-modern/)
22. Tsung-Yi Lin, Priya Goyal, Ross Girshick, Kaiming He, and Piotr Dollár. 2017. [Focal loss for dense object detection](https://md.research.elamur.ai/free/en/papers/lin-2017-focal-loss-dense-object-detection/). In Proceedings of the IEEE International Conference on Computer Vision (ICCV), pages 2980– 2988. IEEE.
23. Manzil Zaheer, Guru Guruganesh, Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, and Amr Ahmed. 2021. [Big bird: Transformers for longer sequences](https://md.research.elamur.ai/free/en/papers/zaheer-2020-big-bird-transformers-longer-sequences/).

*Updated 2 November 2026*