@inproceedings{colak-2026-idiomranker,
    title = "{I}diom{R}anker-{X} at {MWE}-2026 {A}d{MIR}e 2: Multilingual Idiom-Image Alignment via Low-Rank Adaptation of Cross-Encoders",
    author = "Colak, Mehmet Utku",
    editor = {Ojha, Atul Kr.  and
      Mititelu, Verginica Barbu  and
      Constant, Mathieu  and
      Stoyanova, Ivelina  and
      Do{\u{g}}ru{\"o}z, A. Seza  and
      Rademaker, Alexandre},
    booktitle = "Proceedings of the 22nd Workshop on Multiword Expressions ({MWE} 2026)",
    month = mar,
    year = "2026",
    address = "Rabat, Marocco",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-eacl/2026.mwe-1.16/",
    pages = "134--138",
    ISBN = "979-8-89176-363-0",
    abstract = "This paper describes the system submitted for the $\textbf{MWE 2026 Shared Task}$ (AdMIRe 2.0 Subtask A). The submission focused on a text-centric approach, reframing the idiom-image alignment task as a sentence-pair classification problem using $\textbf{mBERT}$ (Multilingual BERT). The submitted system relied on full fine-tuning using only the English training data, achieving a Top-1 Accuracy of approximately $\textbf{0.30}$ on the blind test set. Following the evaluation phase, significant limitations were identified in the cross-lingual generalization of the base model. In a post-evaluation study, the backbone was upgraded to $\textbf{XLM-RoBERTa-Large-XNLI}$, incorporating $\textbf{Low-Rank Adaptation (LoRA)}$ and utilizing the full multilingual dataset with hard negative mining. These improvements boosted the accuracy to $\textbf{0.41}$, demonstrating the necessity of NLI-specific pre-training and parameter-efficient tuning for MWE-aware multimodal tasks."
}