English(EN) Quantiles, Not Points: Streaming Quantile Matching for Ordinal-Aware LLM-as-a-Judge Calibration

新的 LLM 校准方法尊重分数分布

作者 PulseAugur 编辑部 · [1 个来源] · 2026-11-04 05:00

研究人员推出了一种新颖的 LLM-as-a-Judge 分数校准方法——流式分位数匹配 (SQM)。与传统的点校准不同，SQM 将 LLM 分数视为分布而非孤立点，从而更好地与人类评分保持一致。这种方法尊重人类锚定集的序数性，并在各种模型和数据集上实现了卓越的分布保真度和整体排名。SQM 还显示出缩小较小和较大 LLM 裁判之间误差差距的潜力，为更有效的 LLM 评估铺平了道路。 AI

影响这种新的校准方法有望带来更准确、更高效的 LLM 评估，提高 LLM-as-a-Judge 系统的可靠性。

排序理由关于 LLM 评估新方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.CL 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 40 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 关于 LLM 评估新方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, model release; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.CL TIER_1 English(EN) · Ilija Subasic, Andrew Rabinovich, Zhao Chen · 2026-11-04 05:00

分位数而非点：用于序数感知的大语言模型即裁判校准的流式分位数匹配

arXiv:2611.00447v1 Announce Type: cross Abstract: LLM-as-a-judge scores are often distributionally shifted from human ratings on any given evaluation task, and post-hoc calibration (mapping LLM to human scores in a small labeled anchor set) has become a popular method to correct …

本文如何被排名

报道来源 [1]

分位数而非点：用于序数感知的大语言模型即裁判校准的流式分位数匹配

相关实体

相关话题