MRL 2026
Shared Task

Multilingual Representation Learning Workshop @ EMNLP 2026

Following the structure of last year's shared task, this year's MRL Workshop Shared Task to create a community-written multilingual benchmark. The task will be question answering, similar to ECLeKTic. This year's task will be multimodal and will include audio and text formats.

Register Interest Join Discord GitHub 🤗 Hugging Face

Overview

Last year, over 350 contributors from around the world came together to create Global PIQA. This dataset evaluates culturally specific commonsense reasoning performance for over 100 language varieties. All contributors were invited to be authors on the dataset.

The 2026 Shared Task will again involve making a community-led dataset of language- and culture-specific QA items. All accepted submission authors will be included on the resulting benchmark paper. We especially encourage participation from speakers of low-resource languages and non-standard or non-prestige varieties.

2025 Results Now Available

Global PIQA v0.1 was released in October 2025. Explore the dataset on Hugging Face or read the preprint. See the 2025 Shared Task page for full details.

Preprint 🤗 Dataset 2025 Task →

Task Format

The full description of the task format can be found here. Contributors should submit question-answer pairs for language- and culture-specific knowledge.

First, contributors will create text question-answer pairs. Example of a text question-answer pair

After being approved by the organizers, contributors will receive instructions about recording and submitting audio recordings of the items they wrote. Example of audio recording submission

Here is an example of a potential question:

{
  "question": "In deutschsprachigen Filmen, wer synchronisiert Brad Pitt?",
  "answer": "Tobias Meister",
  "translated_question": "In German-speaking films, who dubs Brad Pitt?",
  "translated_answer": "Tobias Meister",
  "country": "Germany",
  "language": "deu_latn",
  "url": "https://de.wikipedia.org/wiki/Tobias_Meister"
}

This format is based on the ECLeKTic dataset, which you can consult for further examples.

Important Dates

All deadlines are Anywhere on Earth (AoE). Dates subject to change — register your interest to receive updates.

June 8th 2026
Call for participation released — task specifications and submission guidelines published.
July 10th 2026
Information Session — online event to find out how to participate and ask questions.
August 1st 2026
Dataset submission deadline — submit your dataset and dataset description.
September 2026
Decision notification — acceptance decisions sent to participants.
Early October 2026
Benchmark paper publication — organizers work with authors to compile and publish the final dataset and paper.
October 24-29 2026
MRL Workshop @ EMNLP 2026 — shared task results presented at the workshop.

Interested in contributing?

Fill out the interest form and we'll be in touch.

Register Interest

Contact

Questions about the shared task? Reach out to the organizers.

Email: mrlbenchmarks@gmail.com
Twitter/X: @mrl_workshop
Bluesky: @mrl-workshop.bsky.social
Discord: Join here!