For Task 3, could you please clarify how the “single unified conditional model” requirement is interpreted? Specifically, is this requirement primarily about using one unified inference framework consistently across all field-strength pairs and modalities, or are there additional restrictions on the internal model design? Thank you for the clarification.

Created by Yuehan Wu Yuehan99
For the detailed interpretation of the single unified conditional model requirement for Task 3, please refer to the Additional Constraints section on the MRIxFields2026 Tasks page: [MRIxFields2026 – Task 3 Additional Constraints](mrixfields.chihucloud.com/2026/tasks.html) In brief, Task 3 requires single jointly trained, parameter-sharing conditional model supporting all source-to-target field transformations. Field-specific, target-specific, or source–target-pair-specific dedicated models/subnetworks selected by a router are not allowed. Please follow the detailed constraints on the website when determining whether your model design is compliant. Best regards, MRIxFields2026 Organizing Committee

Clarification on “single unified conditional model” for Task 3 page is loading…