TTLab at StanceEval-2026: A Cloze-Style Prompting Approach for Arabic-Language Stance Detection (CLASP-Ar)

2026-09-24

AI Summary

This study addresses the high model complexity and poor cross-task transferability of multi-task learning approaches in Arabic stance detection by proposing the CLASP-Ar framework. Departing from conventional multi-task ensemble paradigms, this work reformulates stance detection as a cloze-style masked language modeling task. By integrating prompt engineering with a verbalizer mapping mechanism, the proposed method enables direct prediction over a constrained label space through a single prompt template. This approach significantly simplifies the model architecture while enhancing system applicability and interpretability. Experimental results demonstrate that CLASP-Ar substantially reduces computational complexity while maintaining superior performance on Arabic stance detection benchmarks.

Abstract

Arabic-language stance detection remains challenging, and previous shared-task systems have largely relied on multitask learning and ensembles. While these systems achieve state-of-the-art performance, their applicability and transferability are limited by the additional complexity introduced by multitask learning.To reduce this complexity, we introduce $\texttt{CLASP-Ar}$, which reformulates the task as cloze-style masked language modeling. In this approach, the target, predicted sentiment, and text are combined into a single prompt whose $\texttt{[MASK]}$ prediction is restricted to a verbalizer-constrained label vocabulary.

Problem

• Arabic stance detection
• multitask learning complexity
• transferability
• applicability

Innovation

• Arabic stance detection
• Cloze-style prompting
• Masked language modeling
• Verbalizer
• CLASP-Ar