FAQ
For any questions related to the MultiClinAI shared task, please contact:
- Salvador Lima-López – salvador.limalopez@gmail.com
- Fernando Gallego-Donoso – fgallegodonoso@gmail.com
Q: What is the goal of the MultiClinAI shared task?
The goal of MultiClinAI is twofold:
(i) to evaluate multilingual clinical entity recognition systems across seven languages, and
(ii) to assess automatic methods for generating comparable multilingual clinical corpora through annotation projection.
Participants are required to predict clinical entity annotations and their corresponding character offsets in the provided test datasets.
Q: How do I register?
To participate, you must complete the registration form.
Once your registration is approved, you will be added to a task-specific Google Group, which will be used for all official communications, data releases, and important updates.
Q: The task consists of two sub-tasks. Do I need to participate in both?
No. The two sub-tasks (MultiClinNER and MultiClinCorpus) are independent.
Participants may choose to participate in one or both sub-tasks and may submit results for any subset of the target languages.
Q: Do I need to participate in all languages within a sub-task?
No. Participation is flexible. Teams may choose to participate in only one language within a sub-task, or in any subset of the available languages. Submission to all seven languages is not mandatory.
Q: What exactly do I need to upload?
Participants must upload a single ZIP file to the CodaBench platform. This ZIP file must contain all the run.tsv files corresponding to the submitted runs, following the required submission format described on the Submissions page of the task website.
In addition, teams must send an email to the task organizers including:
- The exact same ZIP file uploaded to CodaBench (as a backup copy).
- A mandatory README file containing a detailed description of the methodology used for each submitted run.
The README must clearly explain the approach followed for each run (e.g., model architecture, training data, external resources, prompting strategy, hyperparameters, or projection method, if applicable).
Q: Should prediction files include headers?
Yes. All .tsv prediction files must include headers with exactly these columns: filename, label, start_span, end_span, text.
Q: How do I submit my results?
Participants must submit their predictions as a ZIP file containing the required output files, following the submission format described on the task website. All official submissions must be uploaded to the CodaBench evaluation platform, which will be used for scoring and ranking.
In addition, teams are strongly encouraged to send a copy of the submitted ZIP file by email to the task organizers as a backup record of the submission. Detailed submission instructions and example ZIP files will be provided before the evaluation phase opens.
Q: Can I submit multiple runs or prediction files?
Yes. Participants may submit multiple runs during the evaluation phase, subject to the submission limits defined for each phase.
All submitted runs must comply with the required submission format.
Q: Can I use additional training data or external resources?
Yes. Participants are allowed to use additional training data, external resources, and pre-trained models.
However, all such resources must be clearly described in the system description paper submitted by the team.
Q: Are all entity types evaluated?
The datasets include annotations for diseases, symptoms, procedures, and medications.
However, only diseases, symptoms, and procedures are evaluated. Medication annotations are released for completeness but are not considered in the evaluation.
Q: Are all languages evaluated jointly or independently?
Each language is evaluated independently.
Participants may submit results for any subset of the seven task languages, and submission for all languages is not mandatory.
Q: Should prediction files include headers?
Yes. All prediction files must include headers, following the format specified in the submission guidelines.
Q: When will the test data be released?
The test datasets will be released via Zenodo, according to the official task schedule.
Dataset links will be announced through the task website and the Google Group.
Q: Is a paper submission required?
Yes. All registered teams are required to submit a system description paper.
To be included in the official workshop proceedings, at least one author must register for and present the work at the #SMM4H-HeaRD 2026 Workshop, which will be held online.