发表机构
Meta Platforms, Inc.（元平台公司）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文提出期望贝叶斯损失（EBL）指标，用于评估中性发布中的下行风险，弥补传统置信区间重叠方法的缺陷，为实验平台提供可调的安全护栏。
AI 中文摘要
使用传统置信区间重叠评估“中性发布”（例如基础设施升级）存在缺陷：在数据稀缺时过于宽松，在数据充足时又过于严格。为解决此问题，本文引入期望贝叶斯损失（EBL），这是一种连续指标，可量化指标退化的概率和预期严重程度。EBL可直接从标准频率派估计中计算，明确惩罚经验噪声和高方差实验。经专家决策验证，EBL为实验平台提供了严格且可调的护栏，将统计安全性与机构风险偏好对齐。
英文摘要
Evaluating "neutral launches" (e.g., infrastructure upgrades) using traditional confidence interval overlap is flawed: it is dangerously permissive with scarce data and excessively restrictive with abundant data. To resolve this, this paper introduces Expected Bayesian Loss (EBL), a continuous metric that quantifies both the probability and expected severity of metric degradation. Computable directly from standard frequentist estimates, EBL explicitly penalizes empirical noise and high-variance experiments. Validated against expert decisions, EBL provides experimentation platforms with a rigorous, tunable guardrail that aligns statistical safety with institutional risk appetite.
