Arabic social media stance detection excels on unseen targets
StanceEval 2026: The Second Stance Detection Shared Task
Computation and Language
Summary
Detecting whether someone supports or opposes a topic in social media posts is hard, especially when topics change. The authors ran a competition where teams built systems to guess stance on Arabic tweets about different topics. Surprisingly, systems did better on completely new topics than on related ones. This might be because some topics are more strongly debated or tricky to understand because of dialect or sarcasm. Various advanced language AI methods were used, improving accuracy a lot compared to basic models.
What this means in practice
- •For social media monitoring teams: Identify user opinions on emerging topics for timely brand or public sentiment analysis across languages and domains.
- •For customer feedback analysts: Automatically classify customer stance in feedback about new products even when training data is from related but different products.
Authors
Rasha Albalawi, Nuha Albadi, Hamzah Luqman, Asma Yamani, Maram Kurdi, Saad Ezzini, Ahmed Ashraf, Maged Al-Shaibani, Nora Alturayeif
Abstract
StanceEval 2026 is the second edition of the StanceEval shared task series on stance detection in Arabic social media text. Stance detection aims to identify a writer's stance toward a given topic. Given a tweet and a target, participating systems must determine whether the writer's stance is Favor, Against, or None. This edition focuses on cross-target generalization across two distinct evaluation tracks: Track 1 evaluates thematically related cross-target transfer (testing on Women Driving, related to Women Empowerment from training data), while Track 2 evaluates cross-domain transfer to completely unseen targets (E-Cars and Trimester System). The shared task attracted 80 registered teams from 12 countries. During the evaluation phase, 30 unique teams submitted entries, with 21 teams officially ranked in Track 1 and 13 in Track 2 following validation filtering, and 20 teams submitting system-description papers. Participating teams employed diverse methodologies, including fine-tuned pretrained language models, prompt-based and retrieval-augmented large language models (LLMs), fine-tuned LLMs, and hybrid cascades. Top systems achieved impressive $F_{avg2}$ scores of 0.8994 on Track 1 and 0.9400 on Track 2, substantially outperforming the strongest baselines (0.7366 and 0.7475, respectively), where $F_{avg2}$ denotes the macro-averaged F1 score over the Favor and Against classes. Counterintuitively, performance on the unseen targets was higher than on the related target, a disparity could be driven by extreme target polarization, class imbalance, and dialectal or sarcastic nuance across topics.