English(EN) LeakScale: Estimating the Causal Effect of Benchmark Exposure

新框架LeakScale量化AI基准污染影响

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-26 04:00

研究人员开发了一个名为LeakScale的新框架，用于量化基准污染对AI模型产生的因果影响。与以往关注评估数据是否出现在训练集中的方法不同，LeakScale衡量了这种数据暴露对模型性能的实际影响程度。在不同领域和模型家族的实验表明，基准暴露持续提高了准确性，影响范围从3.22到10.66个百分点。这种新方法能够更精确地理解训练数据泄露如何影响基准分数。 AI

影响提供了一种更准确的方法来评估AI模型性能，并理解模型的真实能力。

排序理由学术论文，介绍了一种评估AI基准污染的新方法论。[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: 17 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 学术论文，介绍了一种评估AI基准污染的新方法论。[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。

报道来源 [1]

arXiv cs.CL TIER_1 English(EN) · Divyansh Singh · 2026-10-26 04:00

LeakScale：估算基准测试暴露的因果效应

arXiv:2609.27176v3 Announce Type: replace Abstract: Benchmark contamination is usually treated as a provenance question: did evaluation material appear in training? Provenance cannot supply the quantity needed to interpret a benchmark score: how much exposure changed it. We intro…

本文如何被排名

报道来源 [1]

LeakScale：估算基准测试暴露的因果效应

相关实体

相关话题