William Cheng, Mijung Park, Dongjun Kim · 2026 · arXiv
DOI: 10.48550/arXiv.2610.27710 · open-access PDF
Los modelos de difusión han surgido como la base de los sistemas generativos modernos, pero su alta capacidad de memorización plantea preocupaciones de privacidad. Mientras que el entrenamiento diferencial privado (DP) proporciona garantías formales, viene a un coste en calidad de muestra. En este trabajo, estudiamos una pregunta complementaria: ¿cuánta pérdida de privacidad se incurre durante el proceso de muestreo mismo? Analizar directamente a los detonadores neuronales modernos es desafiante porque su sensibilidad a un solo ejemplo de entrenamiento generalmente es intractable. Para superar este desafío, construimos y analizamos un experimento empírico de privacidad.
Machine translation (Meta M2M100); the English original follows.
Diffusion models have emerged as the foundation of modern generative systems, yet their high memorization capacity raises privacy concerns. While differentially private (DP) training provides formal guarantees, it comes at a cost in sample quality. In this work, we study a complementary question: how much privacy loss is incurred during the sampling process itself? Directly analyzing modern neural denoisers is challenging because their sensitivity to a single training example is generally intractable. To overcome this challenge, we construct and analyze a clipped privacy-accountable empirical
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.27710, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange