The participants are asked to identify narratively similar stories. The narrative similarity is defined by three core similarity components: the abstract theme, the course of action, and the outcomes of a story.
We primarily focus on tackling Track A, where the system gets a triple consisting of an anchor story and two choices: A and B. The system is asked to identify which of the two choices is more similar.
More information can be found on the Website of Task4
SemEval2026-Task_4-sample-v1: Sample data (39 items)
SemEval2026-Task_4-dev-v1: Development data (200 items, we developed our methods using this dataset)
SemEval2026-Task_4-test-v1: Test data (400 items)
trackAcode: Our code for Track A
papers_and_feedback: Our draft, the feedback from the SemEval community, and our camera-ready paper
results: Sample output
Yue Yu, Wanzhao Zhang