In this year’s issue of the ARDUOUS workshop, we want to put the focus on the challenges and impact of annotation in the era of large language models (LLM) and foundation models (FM). LLMs and FMs need massive amounts of data in order to be trained, while high-quality annotation does not scale up, it is expensive and slow. What is more, small errors in the annotation could potentially propagate as the models are used for various tasks. LLMs and FMs are increasingly used for pre-labelling, rapid labelling or generation of synthetic data and annotation. This speeds up the annotation process but in the same time injects bias and errors in the annotated data as models train on their own mistakes and in the same time reduce the diversity of interpretation. Another big challenge is that LLMs and FMs are meant for long-term reusability across domains but annotations are frozen in time and reflect the norms, laws and facts of this time. Finally, traditional measures for evaluating the quality of the annotation might potentially no longer be adequate as we do not even know which annotations matter the most during model training and which errors could have a catastrophic impact.

We aim to bring together researchers from the AI community who work on topics addressing the challenges involved in producing reliable and quality-assured annotation in the era of LLMs and FMs. We encourage researchers within the community to share their experience of

• producing annotation for the training, fine-tuning or adapting LLMs and FMs,
• the role and impact of annotations in designing and validating AI applications or training large models,
• the process of labelling, and the requirements to produce high quality annotations for diverse settings and tasks,
• ensuring the maintenance of the labels over time,
• innovative tools, interfaces and automated methods for annotating data,
• methods for standardisation and normalisation in annotation practices,
• evaluation methods for the quality assurance of the annotation in the era of LLMs and FMs and
• novel topics and approaches in this field.

Submission Guidelines

The proceedings will be published in the Communications in Computer Science and Information Science Springer series: https://www.springer.com/series/7899

Format:
For your submission you should use one of the following dedicated templates:

• (.doc format) https://resource-cms.springernature.com/springer-cms/rest/v1/content/19238706/data/v5
• (LaTeX) https://resource-cms.springernature.com/springer-cms/rest/v1/content/19238648/data/v8

The papers should not exceed the following page limits including references:

• Full paper: 12 pages
• Short paper: 8 pages
• Poster and demo paper: 3 to 5 pages

Submission: through the EasyChair submission system at https://easychair.org/conferences/?conf=arduous2026

Review process: the review process will be double blind

Committees

Organising Committee

• Gregory Tourte, University of Oxford, UK
• Kristina Yordanova, University of Greifswald, DE
• Emma Tonkin, University of Bristol, UK

Publication

ARDUOUS 2026 proceedings will be published in the Communications in Computer Science and Information Science Springer series: https://www.springer.com/series/7899

Venue

The 10th International Workshop on Annotation of Real World Data for Artificial Intelligent Systems will be held as part of the German Conference on Artificial Intelligence (KI) 2026 in Bremen, Germany https://ki2026.gi.de/

Contact

All questions about submissions should be emailed to organizers@arduous.eu