Eligibility and registration
Participation in AMPLIFAI requires approval by the organizing team. Participants must provide accurate information regarding their name, organization, nationality, and contact information. The AMPLIFAI organizers reserve the right to approve or deny participation at their discretion.
Data access and restrictions
This challenge uses medical imaging data derived from U.S.-based clinical sources. These data are not publicly distributed. All model evaluation is performed within a secure computing environment controlled by the organizers. Participants will not be granted direct access to the evaluation dataset.
Participants agree that they will not:
- Attempt to access, download, reconstruct, or infer the protected evaluation data
- Attempt to extract sensitive information through model behavior or repeated submissions
- Use the challenge platform to derive patient-level or identifiable information
Submission and execution policy
All submissions:
- Must be self-contained and executable within the provided environment
- Must not attempt to access external systems, open network connections, or write unauthorized outputs
- Must not attempt to reverse engineer the dataset
- Must not perform adversarial attacks aimed at extracting data
The organizers reserve the right to inspect, reject, or terminate any submission that violates these policies.
Output restriction
Participants will only receive aggregate evaluation metrics.
No raw outputs, intermediate data, or derived artifacts (including but not limited to segmentation masks, feature embeddings, or reconstructed images) will be returned.
Monitoring and enforcement
All submissions and activities may be logged and monitored for security and compliance purposes.
Responsible use
Participants agree to use the challenge solely for research and educational purposes.
Publication embargo
We intend to publish a joint challenge paper describing the task, dataset, and results in a peer-reviewed journal. The top 3 teams will be asked to nominate a co-author for this joint publication. A 12-month embargo period will apply only to publications that directly disclose or duplicate challenge-level evaluation results, including rankings or comparative performance on the private held-out test set.
During the embargo period, participating teams retain full ownership of their methods and are free to continue developing, extending, and internally disseminating their work. Publications that substantially extend the submitted method or evaluate it outside the official challenge test set are permitted once the embargo period concludes.
This policy is intended to enable a coherent and archival challenge summary publication without restricting independent methodological research or scientific ownership.
Data license
The training and validation dataset is released under the CC BY-NC-SA license.