Organizing MMCultureQA at SemEval 2027

Created in July 09, 2026

2026

We are happy to share that we will be organizing a SemEval 2027 shared task on Multilingual and Multimodal Culturally Grounded Question Answering.

The task focuses on culturally grounded visual question answering, where a system is given an image and a question and must generate a short, open-ended answer. The question may be presented as speech or text, in multiple languages including Arabic (MSA and dialectal varieties), Bangla, English, Hindi, Assamese, Nepali, Urdu, Italian.

An example from the OASIS dataset is shown below:

The shared task includes two tracks:

  • Task 1: Spoken Visual QA
  • Task 2: Textual Visual QA

More information is available on the official task website: https://mmcultureqa-semeval27.github.io/