English(EN) Single-Query Black-Box Calibration Auditing via Logit Bias

新方法利用 Logit Bias 审计 LLM 校准

作者 PulseAugur 编辑部 · [1 个来源] · 2026-09-07 04:00

研究人员开发了一种新方法，用于审计大型语言模型 (LLM) 在其连续输出概率被隐藏时的校准情况。通过操纵 logit_bias 参数，每个样本的单次查询可用于评估精确的概率阈值。该技术为二元任务引入了一种新颖且一致的真实校准误差估计器，为审计黑盒基础模型提供了一个高效的框架。 AI

影响通过提供一种即使在内部概率被隐藏时也能审计校准的方法，从而能够对 LLM 进行更鲁棒的安全评估。

排序理由该集群包含一篇研究论文，详细介绍了一种审计 LLM 校准的新方法。[lever_c_demoted from research: ic=1 ai=1.0]

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: 32 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇研究论文，详细介绍了一种审计 LLM 校准的新方法。[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, safety; 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。

arXiv cs.LG TIER_1 English(EN) · Roman Plaud, Antoine Saillenfest, Matthieu Labeau, Thomas Bonald, Willem Waegeman · 2026-09-07 04:00

通过 Logit 偏差进行单查询黑盒校准审计

arXiv:2609.05125v1 Announce Type: new Abstract: Evaluating the calibration of Large Language Models (LLMs) is critical for their safe deployment as zero-shot classifiers. Yet, commercial API providers increasingly hide the continuous output probabilities required by standard cali…