{
  "abstract": "In diffusion transformers, low-rank branches can mitigate 4-bit weight--activation (W4A4) post-training quantization (PTQ) loss by decomposing each weight into a low-bit residual and a high-precision low-rank component. Existing low-rank PTQ approaches, however, either optimize low-rank compensation and residual quantization separately, often requiring higher ranks, or rely on second-order weight updates without explicitly modeling activation quantization error, which becomes particularly pronounced under 4-bit quantization. To address these limitations, we present HSVDQuant, a unified framework modeling low-rank-assisted W4A4 PTQ as a coupled calibration problem and deriving optimization-based solvers from the joint objective. Eliminating the output-side low-rank factor yields a deflated Hessian that discounts residual errors already captured by the low-rank component, while an activation-noise surrogate is incorporated to suppress activation quantization error. Across five diffusion backbones, rank-4 HSVDQuant consistently outperforms rank-4 SVDQuant in PSNR and LPIPS. It further surpasses rank-32 SVDQuant on SANA-1.6B, FLUX.1-schnell, and FLUX.1-dev with an $8\\times$ smaller rank and up to $6.25\\times$ faster quantization. Furthermore, on the Qwen3-8B LLM, rank-4 HSVDQuant improves MMLU accuracy from 61.50% to 68.17% over rank-32 SVDQuant. Overall, HSVDQuant achieves better W4A4 performance with substantially lower rank and quantization cost.",
  "abstract_shown": true,
  "affiliations": [],
  "article_type": null,
  "arxiv": "2610.11315",
  "author_total": 3,
  "authors": [
    {
      "name": "Shiwen Wang"
    },
    {
      "name": "Pengxiang Zhao"
    },
    {
      "name": "Xiaoming Yuan"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/34p1fw31vedva.bib",
  "cited_by": 0,
  "doi": "10.48550/arxiv.2610.11315",
  "fulltext_url": null,
  "id": "34p1fw31vedva",
  "issns": [],
  "landing_url": "https://arxiv.org/abs/2610.11315",
  "language": null,
  "licence": "CC-BY-4.0",
  "licence_label": "CC BY 4.0",
  "licence_source": "article",
  "licence_url": "http://creativecommons.org/licenses/by/4.0/",
  "metadata_licence": "CC0-1.0",
  "native_subjects": [
    [
      "arxiv",
      "cs.CV"
    ],
    [
      "arxiv",
      "cs.LG"
    ],
    [
      "arxiv",
      "math.OC"
    ],
    [
      "openalex",
      "1707|Advanced Neural Network Applications"
    ]
  ],
  "openalex": "W7221125114",
  "pdf_url": "https://arxiv.org/pdf/2610.11315",
  "pmcid": null,
  "pmid": null,
  "published": "2026-10-09",
  "publisher": null,
  "ris": "https://research.draconitic.com/paper/34p1fw31vedva.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:28:40Z",
      "native_id": "2610.11315",
      "source": "arxiv",
      "url": "https://arxiv.org/abs/2610.11315"
    }
  ],
  "subjects": [
    {
      "field": "mathematics",
      "origin": "arxiv:math.OC",
      "subfield": "optimisation-control"
    }
  ],
  "title": "Deflating the Hessian: Rank-4 W4A4 Quantization for Multimodal Diffusion Transformers",
  "title_text": "Deflating the Hessian: Rank-4 W4A4 Quantization for Multimodal Diffusion Transformers",
  "type": "preprint",
  "updated_at": "2026-10-30T12:28:40Z",
  "url": "https://research.draconitic.com/paper/34p1fw31vedva",
  "venue": "arXiv",
  "versions": []
}