Overview of the CXR-LT 2026 Challenge: Multi-Center Long-Tailed and Zero Shot Chest X-ray Classification
This addresses the problem of rare disease diagnosis in radiology for clinicians, but it is incremental as it builds on existing benchmarks with new tasks and data.
The paper tackles the challenge of chest X-ray classification under long-tailed distributions and open-world clinical environments by introducing the CXR-LT 2026 benchmark, with top-performing teams achieving an mAP of 0.5854 on known classes and 0.4315 on unseen rare disease classes.
Chest X-ray (CXR) interpretation is hindered by the long-tailed distribution of pathologies and the open-world nature of clinical environments. Existing benchmarks often rely on closed-set classes from single institutions, failing to capture the prevalence of rare diseases or the appearance of novel findings. To address this, we present the CXR-LT 2026 challenge. This third iteration of the benchmark introduces a multi-center dataset comprising over 145,000 images from PadChest and NIH Chest X-ray datasets. The challenge defines two core tasks: (1) Robust Multi-Label Classification on 30 known classes and (2) Open-World Generalization to 6 unseen (out-of-distribution) rare disease classes. We report the results of the top-performing teams, evaluating them via mean Average Precision (mAP), AUROC, and F1-score. The winning solutions achieved an mAP of 0.5854 on Task 1 and 0.4315 on Task 2, demonstrating that large-scale vision-language pre-training significantly mitigates the performance drop typically associated with zero-shot diagnosis.