---
title: "ITUNLP at MWE-2026 AdMIRe 2 — Site 2026"
description: "This paper describes the ITUNLP entry to AdMIRe 2, a shared task that tests whether systems can interpret idioms across many languages using both text…"
authors: ["Atakan Site", "Oğuz Ali Arslan", "Gülşen Eryiğit"]
year: 2026
venue: "Proceedings of the 22nd Workshop on Multiword Expressions (MWE 2026)"
source: "https://aclanthology.org/2026.mwe-1.30/"
doi: "10.18653/v1/2026.mwe-1.30"
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/papers/site-2026-itunlp-mwe-2026-admire-2/"
---
[Elamur Research](https://md.research.elamur.ai/) · [this page in HTML](https://research.elamur.ai/free/papers/site-2026-itunlp-mwe-2026-admire-2/)

[home](https://md.research.elamur.ai/) / [papers](https://md.research.elamur.ai/free/papers/) / [2026](https://md.research.elamur.ai/free/papers/2026/page/3/) / ITUNLP at MWE-2026 AdMIRe 2: A Zero-Shot LLM Pipeline for Multimodal Idiom Understanding and Ranking

# ITUNLP at MWE-2026 AdMIRe 2: A Zero-Shot LLM Pipeline for Multimodal Idiom Understanding and Ranking

**Authors:** Atakan Site, Oğuz Ali Arslan, [Gülşen Eryiğit](https://md.research.elamur.ai/free/authors/gulsen-eryi-it/) · **Year:** 2026 · **Venue:** Proceedings of the 22nd Workshop on Multiword Expressions (MWE 2026) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.mwe-1.30/) · [PDF](https://aclanthology.org/2026.mwe-1.30.pdf) · [DOI 10.18653/v1/2026.mwe-1.30](https://doi.org/10.18653/v1/2026.mwe-1.30) · [OpenAlex](https://openalex.org/W7140101444)

**Citations:** 0 citations (openalex-snapshot · 2026-10-14) · 1 in our library, of the 140673 papers whose bibliographies we read

## Summary
This paper describes the ITUNLP entry to AdMIRe 2, a shared task that tests whether systems can interpret idioms across many languages using both text and images. For each potentially idiomatic expression (PIE) in a context sentence, systems must order five candidate images by how well they reflect the intended meaning, which can be literal or figurative. The task covers 15 languages and has two tracks. In one, images are represented only by their captions. In the other, the actual images are also available. The team built a pipeline around large vision–language models that needs no training. A five-step chain-of-thought prompt first decides whether the PIE is literal or idiomatic, then checks captions and images against that reading and returns a ranking as JSON. Gemini 2.5 Pro is the main model, and GPT-5.1 takes over when safety filters block an answer. Light post-processing repairs inconsistent rankings. The system placed first overall. Its results suggest that careful zero-shot prompting gives a strong baseline for grounding idioms across languages, although idiomatic senses remain harder to handle than literal ones.

## Key points
- **Overall results:** the text-only track reached 55.9% Top-1 accuracy (NDCG@5 0.831), and the multimodal track reached 60.1% (NDCG@5 0.849). No task-specific fine-tuning was used, and the system ranked 1st by average across tracks.
- **Variation by language:** Brazilian Portuguese scored best (78.9% text-only, 85.5% multimodal), and Ecuadorian Spanish scored worst (25.0% / 27.1%). Adding images helped in 14 of 15 languages, with Turkish gaining most (50.5% → 67.8%). Igbo was the only language where images hurt, by about 4 points.
- **Literal vs. idiomatic gap:** literal uses scored 61.1% versus 51.3% for idiomatic uses in text-only. With images, the scores were 66.9% versus 54.8%, so the gap slightly widened.
- **Model comparison:** Gemini 2.5 Pro beat GPT-5.1 on both tracks, by 55.9% vs. 55.0% on text-only and 59.8% vs. 57.3% on multimodal. Gemini also benefited more from visual input.
- **Resources:** the code is released on GitHub (oguzaliarslan/idiom-nlp), and the dataset is described in Torunoğlu-Selamet et al. (2026). Ukrainian is not among the 15 covered languages.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

Fields: [Linguistics & languages](https://md.research.elamur.ai/free/shelves/linguistics/)

## Cite

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{site2026itunlp,
  title = {{ITUNLP at MWE-2026 AdMIRe 2: A Zero-Shot LLM Pipeline for Multimodal Idiom Understanding and Ranking}},
  author = {Site, Atakan and Arslan, Oğuz Ali and Eryiğit, Gülşen},
  booktitle = {Proceedings of the 22nd Workshop on Multiword Expressions (MWE 2026)},
  year = {2026},
  doi = {10.18653/v1/2026.mwe-1.30},
  url = {https://doi.org/10.18653/v1/2026.mwe-1.30}
}
```

- [← previous Ithaca Revisited: Benchmarking a Domain-Specific Model for Epigraphy in the Age of LLMs](https://md.research.elamur.ai/free/papers/locaputo-2026-ithaca-revisited-benchmarking-domain-specific/)
- [next → J-CHAT: Japanese Large-scale Spoken Dialogue Corpus for Spoken Dialogue Language Modeling](https://md.research.elamur.ai/free/papers/nakata-2026-j-chat-japanese-large-scale/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/papers/2026/page/3/)

## References

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

1. (MWE 2026), Rabat, Morocco. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=%28MWE+2026%29%2C+Rabat%2C+Morocco%2E+Association+for+Computational+Linguistics%2E)
2. Tuhin Chakrabarty, Arkadiy Saakyan, Debanjan Ghosh, and Smaranda Muresan. 2022. FLUTE: Figurative language understanding through textual explanations. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 7139–7159, Abu Dhabi, United Arab Emirates. Associat… [find](https://scholar.google.com/scholar?q=Tuhin+Chakrabarty%2C+Arkadiy+Saakyan%2C+Debanjan+Ghosh%2C+and+Smaranda+Muresan%2E+2022%2E+FLUTE%3A+Figurative+language+understanding+through+textual+explanations%2E+In+Proceedings+of+the+2022+Conference+on+Empirica)
3. Gheorghe Comanici, Eric Bieber, Mike Schaekermann, Ice Pasupat, Noveen Sachdeva, Inderjit Dhillon, Marcel Blistein, Ori Ram, Dan Zhang, Evan Rosen, Luke Marris, Sam Petulla, Colin Gaffney, Asaf Aharoni, Nathan Lintz, Tiago Cardal Pais, Henrik Jacobsson, Idan Szpektor, Nan-Jiang Jiang, and 3416 other… [find](https://scholar.google.com/scholar?q=Gheorghe+Comanici%2C+Eric+Bieber%2C+Mike+Schaekermann%2C+Ice+Pasupat%2C+Noveen+Sachdeva%2C+Inderjit+Dhillon%2C+Marcel+Blistein%2C+Ori+Ram%2C+Dan+Zhang%2C+Evan+Rosen%2C+Luke+Marris%2C+Sam+Petulla%2C+Colin+Gaffney%2C+Asaf+Aharon) [our card](https://md.research.elamur.ai/free/papers/comanici-2025-gemini-2-5-pushing-frontier/)
4. Paul Cook, Afsaneh Fazly, and Suzanne Stevenson. 2008. The vnc-tokens dataset. [find](https://scholar.google.com/scholar?q=Paul+Cook%2C+Afsaneh+Fazly%2C+and+Suzanne+Stevenson%2E+2008%2E+The+vnc%2Dtokens+dataset%2E)
5. Doğukan Arslan, Hüseyin Anıl Çakmak, Gulsen Eryigit, and Joakim Nivre. 2025. Using LLMs to advance idiom corpus construction. In Proceedings of the 21st Workshop on Multiword Expressions (MWE 2025), pages 21–31, Albuquerque, New Mexico, U.S.A. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Dog%CC%86ukan+Arslan%2C+H%C3%BCseyin+An%C4%B1l+%C3%87akmak%2C+Gulsen+Eryigit%2C+and+Joakim+Nivre%2E+2025%2E+Using+LLMs+to+advance+idiom+corpus+construction%2E+In+Proceedings+of+the+21st+Workshop+on+Multiword+Expressions+%28MWE+2025%29%2C+)
6. Francesca De Luca Fornaciari, Begoña Altuna, Itziar Gonzalez-Dios, and Maite Melero. 2024. A hard nut to crack: Idiom detection with conversational large language models. In Proceedings of the 4th Workshop on Figurative Language Processing (FigLang 2024), pages 35–44, Mexico City, Mexico (Hybrid). A… [find](https://scholar.google.com/scholar?q=Francesca+De+Luca+Fornaciari%2C+Bego%C3%B1a+Altuna%2C+Itziar+Gonzalez%2DDios%2C+and+Maite+Melero%2E+2024%2E+A+hard+nut+to+crack%3A+Idiom+detection+with+conversational+large+language+models%2E+In+Proceedings+of+the+4th+Wor) [our card](https://md.research.elamur.ai/free/papers/fornaciari-2024-hard-nut-crack-idiom-detection/)
7. Doğukan Arslan, Rodrigo Wilkens, Wei He, Dilara Torunoğlu-Selamet, Thomas Pickard, Aline Villavicencio, Adriana S. Pagano, and Gülşen Eryiğit. 2026. MWE-2026 Shared Task 2: AdMIRe 2 - Advancing Multimodal Idiomaticity Representation. In Proceedings of the 22nd Workshop on Multiword Expressions H… [find](https://scholar.google.com/scholar?q=Dog%CC%86ukan+Arslan%2C+Rodrigo+Wilkens%2C+Wei+He%2C+Dilara+Torunog%CC%86lu%2DSelamet%2C+Thomas+Pickard%2C+Aline+Villavicencio%2C+Adriana+S%2E+Pagano%2C+and+G%C3%BCls%CC%A7en+Eryig%CC%86it%2E+2026%2E+MWE%2D2026+Shared+Task+2%3A+AdMIRe+2+%2D+Advancing+Mu) [our card](https://md.research.elamur.ai/free/papers/haagsma-2020-magpie-large-corpus-potentially-idiomatic/)
8. Kai Golan Hashiloni, Ofri Hefetz, and Kfir Bar. 2025. Easy as PIE? identifying multi-word expressions with LLMs. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 23782–23801, Suzhou, China. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Kai+Golan+Hashiloni%2C+Ofri+Hefetz%2C+and+Kfir+Bar%2E+2025%2E+Easy+as+PIE%3F+identifying+multi%2Dword+expressions+with+LLMs%2E+In+Proceedings+of+the+2025+Conference+on+Empirical+Methods+in+Natural+Language+Processi)
9. Paria Khoshtab, Danial Namazifard, Mostafa Masoudi, Ali Akhgary, Samin Mahdizadeh Sani, and Yadollah Yaghoobzadeh. 2025. Comparative study of multilingual idioms and similes in large language models. In Proceedings of the 31st International Conference on Computational Linguistics, pages 8680–8698, A… [find](https://scholar.google.com/scholar?q=Paria+Khoshtab%2C+Danial+Namazifard%2C+Mostafa+Masoudi%2C+Ali+Akhgary%2C+Samin+Mahdizadeh+Sani%2C+and+Yadollah+Yaghoobzadeh%2E+2025%2E+Comparative+study+of+multilingual+idioms+and+similes+in+large+language+models%2E+) [our card](https://md.research.elamur.ai/free/papers/khoshtab-2025-comparative-study-multilingual-idioms-similes/)
10. Jisu Kim, Youngwoo Shin, Uiji Hwang, Jihun Choi, Richeng Xuan, and Taeuk Kim. 2025. Memorization or reasoning? exploring the idiom understanding of LLMs. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 21689–21710, Suzhou, China. Association for Compu… [find](https://scholar.google.com/scholar?q=Jisu+Kim%2C+Youngwoo+Shin%2C+Uiji+Hwang%2C+Jihun+Choi%2C+Richeng+Xuan%2C+and+Taeuk+Kim%2E+2025%2E+Memorization+or+reasoning%3F+exploring+the+idiom+understanding+of+LLMs%2E+In+Proceedings+of+the+2025+Conference+on+Empir) [our card](https://md.research.elamur.ai/free/papers/kim-2025-memorization-or-reasoning-exploring-idiom/)
11. Frances Laureano De Leon, Harish Tayyar Madabushi, and Mark G. Lee. 2025. Evaluating large language models on multiword expressions in multilingual and code-switched contexts. Preprint, arXiv:2504.20051. [find](https://scholar.google.com/scholar?q=Frances+Laureano+De+Leon%2C+Harish+Tayyar+Madabushi%2C+and+Mark+G%2E+Lee%2E+2025%2E+Evaluating+large+language+models+on+multiword+expressions+in+multilingual+and+code%2Dswitched+contexts%2E+Preprint%2C+arXiv%3A2504%2E200)
12. Harish Tayyar Madabushi, Edward Gow-Smith, Carolina Scarton, and Aline Villavicencio. 2021. Astitchinlanguagemodels: Dataset and methods for the exploration of idiomaticity in pre-trained language models. Preprint, arXiv:2109.04413. [find](https://scholar.google.com/scholar?q=Harish+Tayyar+Madabushi%2C+Edward+Gow%2DSmith%2C+Carolina+Scarton%2C+and+Aline+Villavicencio%2E+2021%2E+Astitchinlanguagemodels%3A+Dataset+and+methods+for+the+exploration+of+idiomaticity+in+pre%2Dtrained+language+mod) [our card](https://md.research.elamur.ai/free/papers/madabushi-2021-astitchinlanguagemodels-dataset-methods-exploration-idiomaticity/)
13. Maggie Mi, Aline Villavicencio, and Nafise Sadat Moosavi. 2025. Rolling the DICE on idiomaticity: How LLMs fail to grasp context. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 7314–7332, Vienna, Austria. Association for Comp… [find](https://scholar.google.com/scholar?q=Maggie+Mi%2C+Aline+Villavicencio%2C+and+Nafise+Sadat+Moosavi%2E+2025%2E+Rolling+the+DICE+on+idiomaticity%3A+How+LLMs+fail+to+grasp+context%2E+In+Proceedings+of+the+63rd+Annual+Meeting+of+the+Association+for+Compu)
14. Arkadiy Saakyan, Shreyas Kulkarni, Tuhin Chakrabarty, and Smaranda Muresan. 2025. Understanding figurative meaning through explainable visual entailment. Preprint, arXiv:2405.01474. [find](https://scholar.google.com/scholar?q=Arkadiy+Saakyan%2C+Shreyas+Kulkarni%2C+Tuhin+Chakrabarty%2C+and+Smaranda+Muresan%2E+2025%2E+Understanding+figurative+meaning+through+explainable+visual+entailment%2E+Preprint%2C+arXiv%3A2405%2E01474%2E)
15. Prateek Saxena and Soma Paul. 2020. Epie dataset: A corpus for possible idiomatic expressions. Preprint, arXiv:2006.09479. — [arXiv:2006.09479](https://arxiv.org/abs/2006.09479) [find](https://scholar.google.com/scholar?q=Prateek+Saxena+and+Soma+Paul%2E+2020%2E+Epie+dataset%3A+A+corpus+for+possible+idiomatic+expressions%2E+Preprint%2C+arXiv%3A2006%2E09479%2E+%E2%80%94+%5BarXiv%3A2006%2E09479%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2006%2E09479%29)
16. Kevin Stowe, Prasetya Utama, and Iryna Gurevych. 2022. IMPLI: Investigating NLI models’ performance on figurative language. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 5375–5388, Dublin, Ireland. Association for Computatio… [find](https://scholar.google.com/scholar?q=Kevin+Stowe%2C+Prasetya+Utama%2C+and+Iryna+Gurevych%2E+2022%2E+IMPLI%3A+Investigating+NLI+models%E2%80%99+performance+on+figurative+language%2E+In+Proceedings+of+the+60th+Annual+Meeting+of+the+Association+for+Computation) [our card](https://md.research.elamur.ai/free/papers/stowe-2022-impli-investigating-nli-models-performance/)
17. Harish Tayyar Madabushi, Edward Gow-Smith, Marcos Garcia, Carolina Scarton, Marco Idiart, and Aline Villavicencio. 2022. SemEval-2022 task 2: Multilingual idiomaticity detection and sentence embedding. In Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022), pages 107… [find](https://scholar.google.com/scholar?q=Harish+Tayyar+Madabushi%2C+Edward+Gow%2DSmith%2C+Marcos+Garcia%2C+Carolina+Scarton%2C+Marco+Idiart%2C+and+Aline+Villavicencio%2E+2022%2E+SemEval%2D2022+task+2%3A+Multilingual+idiomaticity+detection+and+sentence+embedding) [our card](https://md.research.elamur.ai/free/papers/madabushi-2022-semeval-2022-task-2-multilingual/)
18. Simone Tedeschi, Federico Martelli, and Roberto Navigli. 2022. ID10M: Idiom identification in 10 languages. In Findings of the Association for Computational Linguistics: NAACL 2022, pages 2715–2726, Seattle, United States. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Simone+Tedeschi%2C+Federico+Martelli%2C+and+Roberto+Navigli%2E+2022%2E+ID10M%3A+Idiom+identification+in+10+languages%2E+In+Findings+of+the+Association+for+Computational+Linguistics%3A+NAACL+2022%2C+pages+2715%E2%80%932726%2C+S) [our card](https://md.research.elamur.ai/free/papers/tedeschi-2022-id10m-idiom-identification-10-languages/)
19. Dilara Torunoğlu-Selamet, Dogukan Arslan, Rodrigo Wilkens, Wei He, Doruk Eryiğit, Thomas Pickard, Adriana S. Pagano, Aline Villavicencio, Gülşen Eryiğit, Ágnes Abuczki, Aida Cardoso, Alesia Lazarenka, Dina Almassova, Amalia Mendes, Anna Kanellopoulou, Antoni Brosa-Rodríguez, Baiba Saulite, Beata… [find](https://scholar.google.com/scholar?q=Dilara+Torunog%CC%86lu%2DSelamet%2C+Dogukan+Arslan%2C+Rodrigo+Wilkens%2C+Wei+He%2C+Doruk+Eryig%CC%86it%2C+Thomas+Pickard%2C+Adriana+S%2E+Pagano%2C+Aline+Villavicencio%2C+G%C3%BCls%CC%A7en+Eryig%CC%86it%2C+%C3%81gnes+Abuczki%2C+Aida+Cardoso%2C+Alesia+Lazare) [our card](https://md.research.elamur.ai/free/papers/selamet-2026-parallel-cross-lingual-benchmark-multimodal/)
20. Özge Umut, Atakan Site, Doğukan Arslan, and Gülşen Eryiğit. 2025. Exploring Turkish idiomaticity with large language models. In 2025 10th International Conference on Computer Science and Engineering (UBMK), pages 533–538. [find](https://scholar.google.com/scholar?q=%C3%96zge+Umut%2C+Atakan+Site%2C+Dog%CC%86ukan+Arslan%2C+and+G%C3%BCls%CC%A7en+Eryig%CC%86it%2E+2025%2E+Exploring+Turkish+idiomaticity+with+large+language+models%2E+In+2025+10th+International+Conference+on+Computer+Science+and+Engineerin)
21. OpenAI. 2025. Gpt-5.1 instant and gpt-5.1 thinking system card. [https://cdn.openai.com/pdf/](https://cdn.openai.com/pdf/) 4173ec8d-1229-47db-96de-06d87147e07e/5_1_ system_card.pdf. Technical report. [↗](https://cdn.openai.com/pdf/)
22. Thomas Pickard, Aline Villavicencio, Maggie Mi, Wei He, Dylan Phelps, and Marco Idiart. 2025. SemEval-2025 task 1: AdMIRe - advancing multimodal idiomaticity representation. In Proceedings of the 19th International Workshop on Semantic Evaluation (SemEval-2025), pages 2597–2609, Vienna, Austria. Ass… [find](https://scholar.google.com/scholar?q=Thomas+Pickard%2C+Aline+Villavicencio%2C+Maggie+Mi%2C+Wei+He%2C+Dylan+Phelps%2C+and+Marco+Idiart%2E+2025%2E+SemEval%2D2025+task+1%3A+AdMIRe+%2D+advancing+multimodal+idiomaticity+representation%2E+In+Proceedings+of+the+19th) [our card](https://md.research.elamur.ai/free/papers/pickard-2025-semeval-2025-task-1-admire/)
23. Geetanjali Rakshit and Jeffrey Flanigan. 2022. FigurativeQA: A test benchmark for figurativeness comprehension for question answering. In Proceedings of the 3rd Workshop on Figurative Language Processing (FLP), pages 160–166, Abu Dhabi, United Arab Emirates (Hybrid). Association for Computational Li… [find](https://scholar.google.com/scholar?q=Geetanjali+Rakshit+and+Jeffrey+Flanigan%2E+2022%2E+FigurativeQA%3A+A+test+benchmark+for+figurativeness+comprehension+for+question+answering%2E+In+Proceedings+of+the+3rd+Workshop+on+Figurative+Language+Process)
24. Ron Yosef, Yonatan Bitton, and Dafna Shahaf. 2023. IRFL: Image recognition of figurative language. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 1044–1058, Singapore. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Ron+Yosef%2C+Yonatan+Bitton%2C+and+Dafna+Shahaf%2E+2023%2E+IRFL%3A+Image+recognition+of+figurative+language%2E+In+Findings+of+the+Association+for+Computational+Linguistics%3A+EMNLP+2023%2C+pages+1044%E2%80%931058%2C+Singapore%2E)
25. Runyang You, Xinyue Mei, and Mengyuan Zhou. 2025. PALI-NLP at SemEval 2025 task 1: Multimodal idiom recognition and alignment. In Proceedings of the 19th International Workshop on Semantic Evaluation (SemEval-2025), pages 1211–1216, Vienna, Austria. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Runyang+You%2C+Xinyue+Mei%2C+and+Mengyuan+Zhou%2E+2025%2E+PALI%2DNLP+at+SemEval+2025+task+1%3A+Multimodal+idiom+recognition+and+alignment%2E+In+Proceedings+of+the+19th+International+Workshop+on+Semantic+Evaluation+)