{
  "abstract": "Speculative sampling accelerates diffusion generation by verifying inexpensive draft states in parallel while preserving the target law. Recent tree-based methods allocate the parallel compute budget more effectively than single-chain drafts, as demonstrated by Diffusion Greedy Rejection Sampling (D-GRS). D-GRS generates $K$ conditionally independent candidates per node, and sequentially tests them in their generation order. Yet the sampled candidates admit an informative ranking without additional target-model evaluations. To exploit this, we introduce Rank-Aware Speculative Sampling (RASS), a verification rule for speculative draft trees based on rank-aware list coupling. RASS orders draft candidates along the proposal-target mean displacement and samples a rank with weights optimized to minimize total variation between the selected-proposal and target laws. Finally, the selected candidate is maximally coupled with the target, with residual correction ensuring exact sampling for any choice of rank weights. We evaluate RASS on a Gaussian-mixture target, unconditional pixel-space generation on FFHQ, conditional generation on CIFAR-10, and latent diffusion with Stable Diffusion 3.5 using COCO2014 prompts. Measured by the ratio of standard to speculative sampling's target-model evaluation counts, RASS improves on D-GRS across the evaluated settings, with gains reaching approximately 20% on CIFAR-10 at matched compute budgets.",
  "abstract_shown": true,
  "affiliations": [],
  "article_type": null,
  "arxiv": "2610.02251",
  "author_total": 5,
  "authors": [
    {
      "name": "Marcello Bullo"
    },
    {
      "name": "Yanxiao Liu"
    },
    {
      "name": "Öykü Sıla Güner"
    },
    {
      "name": "Arpan Mukherjee"
    },
    {
      "name": "Deniz Gündüz"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/3jq0yetyq3tfq.bib",
  "cited_by": 0,
  "doi": "10.48550/arxiv.2610.02251",
  "fulltext_url": null,
  "id": "3jq0yetyq3tfq",
  "issns": [],
  "landing_url": "https://arxiv.org/abs/2610.02251",
  "language": null,
  "licence": "LicenseRef-arXiv-nonexclusive-1.0",
  "licence_label": "arXiv licence",
  "licence_source": "article",
  "licence_url": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
  "metadata_licence": "CC0-1.0",
  "native_subjects": [
    [
      "arxiv",
      "cs.LG"
    ],
    [
      "arxiv",
      "stat.CO"
    ],
    [
      "openalex",
      "2613|Markov Chains and Monte Carlo Methods"
    ]
  ],
  "openalex": "W7220397015",
  "pdf_url": "https://arxiv.org/pdf/2610.02251",
  "pmcid": null,
  "pmid": null,
  "published": "2026-10-01",
  "publisher": null,
  "ris": "https://research.draconitic.com/paper/3jq0yetyq3tfq.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:26:56Z",
      "native_id": "2610.02251",
      "source": "arxiv",
      "url": "https://arxiv.org/abs/2610.02251"
    }
  ],
  "subjects": [
    {
      "field": "mathematics",
      "origin": "arxiv:stat.CO",
      "subfield": "statistics"
    }
  ],
  "title": "Rank-Aware Speculative Sampling for Diffusion Draft Trees",
  "title_text": "Rank-Aware Speculative Sampling for Diffusion Draft Trees",
  "type": "preprint",
  "updated_at": "2026-10-30T12:26:56Z",
  "url": "https://research.draconitic.com/paper/3jq0yetyq3tfq",
  "venue": "arXiv",
  "versions": []
}