System Description Papers
Camera-ready system description papers submitted by participating teams for presentation at ArabicNLP 2026.
NLP/CORE Unranked
ArabicNLP2026-StanceEval 2026
ArabicNLP 2026 Shared Task on Stance Detection
StanceEval-2026 is a shared task focused on Arabic stance detection in social media, hosted as part of ArabicNLP 2026 @ EMNLP in Budapest. Participants develop systems to classify the stance (Favor, Against, None) of Arabic tweets toward specific targets, using the Mawqif-v2 dataset, with tracks for both seen and unseen targets to evaluate generalization.
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Research areas in scope
01
Task Tracks
Seen TargetsUnseen Targets
Policies worth checking twice
- Submissions must be made via Codabench.
- Blind test data is released for submission after registration.
- Evaluation uses Favg2 and Favg3 metrics as defined in prior Arabic stance detection literature.
- Teams must register by the deadline to participate.
- The evaluation script is provided for reproducibility and implementation.
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026