SemEval 2026 — Semantic Evaluation Shared Tasks, San Diego

SemEval is the long-running shared-task workshop of SIGLEX, where research teams compete on standardized datasets to evaluate how well systems handle computational semantics. The 2026 edition — the 20th — is a workshop co-located with ACL 2026 in San Diego in July; its task evaluation windows ran in early 2026 and system papers were due in spring, so the in-person workshop is the culmination where results are presented and compared. Note this is a workshop hosted under a larger conference rather than a standalone event (the exact day within the 2-7 July ACL week is confirmed closer to the date).

For attendees, SemEval is the field's most concrete benchmark of where automatic understanding of meaning actually stands. The 2026 tasks span the topics that matter most right now: multilingual and multicultural problems, including detecting online polarization across languages and dimensional aspect-based sentiment analysis, alongside the workshop's traditional coverage of word-sense, semantic parsing, coreference and sentiment. In an era where large language models are assumed to "understand" text, SemEval's rigorously scored shared tasks provide the adversarial, dataset-grounded reality check — showing concretely which semantic phenomena current systems handle well and which they still fail on, across many languages rather than English alone.