{
  "id": 4049,
  "url": "https://arxiv.org/abs/2605.20282v3",
  "title": "Do Vision Models Truly Forget? New Findings from Representation-Level Certification of Visual Unlearning in Vertical Federated Learning",
  "summary": "Machine unlearning in Vertical Federated Learning (VFL) has attracted growing interest, yet existing methods certify forgetting solely using output-level metrics. We challenge these works by introducing Mirage, a representation-level auditing framework that comprises four complementary diagnostics: Linear probe recovery (LPR), centered kernel alignment (CKA), feature separability scoring, and layer-wise recovery analysis. Extensive experiments across seven datasets and seven baseline methods fol",
  "authors": "Zhenyu Yu, Yangchen Zeng, Chunlei Meng, Guangzhen Yao, Shuigeng Zhou",
  "category": "research",
  "topics": "safety-alignment,healthcare,transparency",
  "orgs": null,
  "regions": null,
  "published_at": "2026-05-19T05:43:57.000Z",
  "fetched_at": "2026-07-14T16:30:41.584Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/4049",
  "original_url": "https://arxiv.org/abs/2605.20282v3",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}