Shared Task

Shared Task on Multi-Domain Document Understanding

 

The Fifth Ukrainian NLP Conference organizes a Shared Task on Multi-Domain Document Understanding. This Shared Task aims to challenge and assess AI capabilities to find the right information in a stack of domain-specific documents and generalize across domains.

 

Important Dates

 

December 1, 2025 — Shared task announcement

December 20, 2025 — Second call for participation

January 10, 2026 — Third call for participation

March 2, 2026 — The Kaggle competition link is sent to the registered participants

March 29, 2026 — Registration deadline

March 30 (11:58 PM GMT +02:00), 2026 — Final submission of system responses

April 2, 2026 — Results of the Shared Task are announced

April 8, 2026 — Shared Task paper due (extended)

April 15, 2026 — Notification of acceptance

April 26, 2026 — Camera-ready Shared Task papers due

May 29–30, 2026 — Conference dates

 

Task Description

 

In this shared task, your goal is to build a solution that, given a set of documents, can (1) find the correct answer to a multiple-choice question and (2) identify the document and page where the answer was found.

 

The competition will run on Kaggle in the code-only mode, which means that you will submit your solution as code, e.g., a Python script or notebook, rather than submitting just test answers. Your solution should be able to run on a single P100 GPU and complete all test sets prediction within 9 hours. The code-only mode will facilitate fair competition.

 

You are free to use any open-source embeddings and large language models of your choice — just keep in mind the compute limitations, which may not allow you to use the larger LLMs.

 

We encourage you to use LLMs specifically trained to work with the Ukrainian language, like LapaLLM and MamayLM.

 

Find all the details at https://github.com/unlp-workshop/unlp-2026-shared-task.

 

Registration

 

Teams that intend to participate should register by filling in this form by March 29, 2026.

 

Publication

 

Participants in the shared task are invited to submit a paper to the UNLP 2026 conference. Please see the Call for Papers page for details. Accepted papers will appear in the ACL anthology and will be presented at a session of UNLP 2026 specially dedicated to the Shared Task.