{
  "abstract": "LLM judges evaluate candidate language models through pairwise comparisons of their responses. Judge-aware Bradley--Terry models aggregate pairwise comparisons by jointly estimating candidate scores and judge-specific sensitivities (how strongly each judge responds to quality differences), often interpreted as reliability. We show that this reading fails when judges prefer a display position. Even under balanced display orders or swap-and-average, omitting position attenuates fitted judge sensitivity by approximately the factor $1-D^2$ near equal quality, where $D$ is the judge's net first-slot preference between equally good responses. We propose OSCAR, which jointly estimates candidate scores, judge sensitivities and judge-specific position, length and family effects, and provides confidence intervals for pairwise win probabilities that account for dependence within prompts. On held-out data, observed attenuation closely matches the prediction $1-D^2$ (correlation of log ratios 0.99). Across four datasets, position adjustment lowers held-out log loss by 14-23% and changes which judges would be dropped for low sensitivity on three datasets. On MT-Bench, it ranks GPT-4 first instead of claude-v1, matching human votes. Simulations with known parameters show improved recovery of judge sensitivities and candidate rankings, with 94.4-95.2% confidence interval coverage. Low estimated sensitivity alone is therefore not evidence of an unreliable judge.",
  "abstract_shown": true,
  "affiliations": [],
  "article_type": null,
  "arxiv": "2609.24128",
  "author_total": 5,
  "authors": [
    {
      "name": "You Liu"
    },
    {
      "name": "Yue Liu"
    },
    {
      "name": "Quanchao Lu"
    },
    {
      "name": "Ergan Shang"
    },
    {
      "name": "Nick Shipilov"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/9tdpexfn7n0d7.bib",
  "cited_by": 0,
  "doi": "10.48550/arxiv.2609.24128",
  "fulltext_url": null,
  "id": "9tdpexfn7n0d7",
  "issns": [],
  "landing_url": "https://arxiv.org/abs/2609.24128",
  "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.LG"
    ],
    [
      "arxiv",
      "stat.AP"
    ],
    [
      "arxiv",
      "stat.ML"
    ],
    [
      "openalex",
      "1710|Information Retrieval and Search Behavior"
    ]
  ],
  "openalex": "W7214003007",
  "pdf_url": "https://arxiv.org/pdf/2609.24128",
  "pmcid": null,
  "pmid": null,
  "published": "2026-10-13",
  "publisher": null,
  "ris": "https://research.draconitic.com/paper/9tdpexfn7n0d7.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:29:26Z",
      "native_id": "2609.24128",
      "source": "arxiv",
      "url": "https://arxiv.org/abs/2609.24128"
    }
  ],
  "subjects": [
    {
      "field": "mathematics",
      "origin": "arxiv:stat.AP",
      "subfield": "statistics"
    }
  ],
  "title": "OSCAR: Order-aware Sensitivity Correction for AI Rankings",
  "title_text": "OSCAR: Order-aware Sensitivity Correction for AI Rankings",
  "type": "preprint",
  "updated_at": "2026-10-30T12:29:26Z",
  "url": "https://research.draconitic.com/paper/9tdpexfn7n0d7",
  "venue": "arXiv",
  "versions": []
}