{
  "abstract": "Feedback noise is pervasive in preference data, arising from subjective human judgments, annotation mistakes, and imperfect LLM judges. Standard reward modeling trained on such feedback yields a biased estimate of the clean risk. Although classical denoising methods are directly applicable to noisy feedback learning, they generally assume a uniform noise ratio across samples. However, this assumption fails for feedback data, where noise is intrinsically sample-wise. To address this challenge, we propose CleanRM, which uses a few clean feedback to calibrate sample-wise noise ratios via optimal transport. Specifically, CleanRM aligns the joint distribution of preferences reverted from noisy feedback with that of clean feedback and then trains the reward model using a noise-corrected surrogate loss. Theoretically, we prove that the resulting objective is unbiased with respect to the clean risk. Extensive experiments demonstrate that CleanRM outperforms state-of-the-art denoising baselines and yields more effective downstream RLHF alignment.",
  "abstract_shown": true,
  "affiliations": [],
  "article_type": null,
  "arxiv": "2610.21138",
  "author_total": 10,
  "authors": [
    {
      "name": "Licheng Pan"
    },
    {
      "name": "Haocheng Yang"
    },
    {
      "name": "Yuzhou Chen"
    },
    {
      "name": "Yichen Sun"
    },
    {
      "name": "Yuan Lu"
    },
    {
      "name": "Yunfei Teng"
    },
    {
      "name": "Jiajun Fan"
    },
    {
      "name": "Guangyi Liu"
    },
    {
      "name": "Zhixuan Chu"
    },
    {
      "name": "Hao Wang"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/b08k0cekex71q.bib",
  "cited_by": 0,
  "doi": "10.48550/arxiv.2610.21138",
  "fulltext_url": null,
  "id": "b08k0cekex71q",
  "issns": [],
  "landing_url": "https://arxiv.org/abs/2610.21138",
  "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.AI"
    ],
    [
      "arxiv",
      "stat.ML"
    ]
  ],
  "openalex": null,
  "pdf_url": "https://arxiv.org/pdf/2610.21138",
  "pmcid": null,
  "pmid": null,
  "published": "2026-10-15",
  "publisher": null,
  "ris": "https://research.draconitic.com/paper/b08k0cekex71q.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:30:00Z",
      "native_id": "2610.21138",
      "source": "arxiv",
      "url": "https://arxiv.org/abs/2610.21138"
    }
  ],
  "subjects": [
    {
      "field": "mathematics",
      "origin": "arxiv:stat.ML",
      "subfield": "statistics"
    }
  ],
  "title": "CleanRM: Reward Modeling from Noisy Feedback with a Few Clean Feedback for RLHF",
  "title_text": "CleanRM: Reward Modeling from Noisy Feedback with a Few Clean Feedback for RLHF",
  "type": "preprint",
  "updated_at": "2026-10-30T12:30:00Z",
  "url": "https://research.draconitic.com/paper/b08k0cekex71q",
  "venue": "arXiv",
  "versions": []
}