Challenge Submission
Docker-based system submission for the MedReason challenge, evaluating MCQ accuracy, open-ended ground truth (GT), and open-ended visual alignment (VA) on hidden datasets.
AI/CORE Unranked
MedReason2026 2026
MICCAI MedReason Challenge 2026
MedReason2026 is a MICCAI 2026 challenge and workshop track focused on evaluating the reliability of medical multimodal large language models (MLLMs) in visual question answering under domain shift, such as low-dose X-ray and 7T brain MRI. The challenge assesses both answer accuracy and reasoning robustness, with an optional archival paper track for participants to publish their methods in Springer LNCS proceedings.
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Individual Method Paper
Archival paper describing the team's method, experiments, and results from their challenge submission, to be published in MICCAI 2026 Springer LNCS proceedings with up to 8 pages of main text and 2 pages of references.
Research areas in scope
01
Challenge Task
Medical multimodal reasoningVisual question answering (VQA) with medical imagingDomain shift in medical imaging (e.g., low-dose X-ray, 7T MRI)Multiple-choice and open-ended question answeringReasoning trace generation
Policies worth checking twice
- Blind review for paper submissions
- One paper per team with maximum three authors
- No rebuttal stage in the review process
- Papers must be fully anonymized during review (no author names, affiliations, or acknowledgments)
- Camera-ready paper due August 24, 2026
- Submission must use the official MICCAI workshop Springer LNCS template
- Paper submission is optional and does not affect challenge ranking or award eligibility
- No dual submission to other venues implied; joint journal paper is separate and not mutually exclusive
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026