This competition evaluations are hosted on the codabench platform. We have created four task-specific competition sites described and linked below. Please make sure that you sign up to the respective codabench for the task you are attempting.

• Task 1: Figure Classification: https://www.codabench.org/competitions/12901/

• Task 2: Data Table Extraction: https://www.codabench.org/competitions/12902/

• Task 3: Summarization: https://www.codabench.org/competitions/12909/

• Task 4: Visual Question Answering: https://www.codabench.org/competitions/12908/

As a overview: 

for your system training please use https://github.com/sciknoworg/ALD-E-ImageMiner/tree/main/icdar2026-competition-data/train and 

https://github.com/sciknoworg/ALD-E-ImageMiner/tree/main/icdar2026-competition-data/dev for model hyperparameter optimization and related model development related tasks. 

To make your submissions in the practice phase on codabench please apply your models on the small test data batch released https://github.com/sciknoworg/ALD-E-ImageMiner/tree/main/icdar2026-competition-data/test. 

Here are the codabench submission preparation guidelines https://github.com/sciknoworg/ALD-E-ImageMiner/tree/main/icdar2026-competition-data/test/submission_guidelines