{
  "id": 748614,
  "ownerSlug": "fabotelli",
  "slug": "pillar0-knee-finetune",
  "title": "Pillar-0 Knee MRI Fine-tune",
  "subtitle": "Pillar-0 BreastMRI foundation model fine-tuned on RSNA 2026 knee MRI",
  "isPrivate": false,
  "description": "Full fine-tune of YalaLab's Pillar-0 (BreastMRI) radiology foundation model on knee MRI from the RSNA 2026 Knee Abnormality Detection competition. Published as an honest cross-anatomy transfer study: 58-study expert holdout macro-AUC 0.797 (frozen probe 0.697; task-specific CNN teachers 0.89-0.90). Strong on Effusion (0.965) and Medial OA (0.941); the domain gap concentrates in ligament/meniscus fine structure. ECL-2.0, attribution to YalaLab Pillar-0.",
  "publishTime": null
}