{
  "id": 172,
  "url": "https://arxiv.org/abs/2607.06609v1",
  "title": "D2PO: Optimizing Diffusion Samplers via Dynamic Preference",
  "summary": "We propose D2PO (Dynamic Direct Preference Optimization), a principled framework for optimizing diffusion sampling policies with respect to timestep schedules and classifier-free guidance (CFG) weights. Our work is motivated by a fundamental limitation of existing student-teacher regression frameworks; low-NFE student samplers are trained to mimic high-NFEteachers, often sacrificing high-frequency texture fidelity while preserving coarse global structures, thereby misaligning the sampler with pe",
  "authors": "Jinkyu Kim, Jinyoung Choi, Bohyung Han",
  "category": "research",
  "topics": "safety-alignment,children-education",
  "orgs": null,
  "regions": null,
  "published_at": "2026-07-07T06:05:42.000Z",
  "fetched_at": "2026-07-14T14:14:19.971Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/172",
  "original_url": "https://arxiv.org/abs/2607.06609v1",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}