{
  "id": 4522,
  "url": "https://arxiv.org/abs/2605.11347v2",
  "title": "Gradient-Free Noise Optimization for Reward Alignment in Generative Models",
  "summary": "Existing reward alignment methods for diffusion and flow models rely on multi-step stochastic trajectories, making them difficult to extend to deterministic generators. A natural alternative is noise-space optimization, but existing approaches require backpropagation through the generator and reward pipeline, limiting applicability to differentiable settings. To address this, here we present ZeNO (Zeroth-order Noise Optimization), a gradient-free framework that formulates noise optimization as a",
  "authors": "Jeongsol Kim, Hongeun Kim, Jian Wang, Jong Chul Ye",
  "category": "research",
  "topics": "safety-alignment",
  "orgs": null,
  "regions": null,
  "published_at": "2026-05-12T00:05:36.000Z",
  "fetched_at": "2026-07-14T16:31:03.580Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/4522",
  "original_url": "https://arxiv.org/abs/2605.11347v2",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}