重新定义不确定性：数据再多也未必能消除的，才是真正的认知不确定性。
Epistemic Uncertainty Is Not the Reducible Kind
- 提出三类不确定性：偶然性、样本可降解认知、机制不可降解认知
- 证明现有度量方法会误判不可减少的不确定性为可减少项
- 适合研究模型可信度与不确定性建模的学者参考
标准预测不确定性分类将认知不确定性定义为可通过收集更多数据消除的部分，其度量通常对应互信息。我们证明该定义与度量在扩展意义上不一致：在特定构造下，度量将全部不确定性归为认知类，但无论增加多少训练数据都无法减少。可降解性实为（不确定性，采集类别）的属性，二分法应拓展为三类：偶然性、样本可降解的认知、机制不可降解的认知。观测价值的精确恒等式表明，在分布内数据不会减少机制不可降解不确定性，且通常会加剧它。目前广泛使用的集成分歧作为认知不确定性的代理，反映的是训练过程而非真实认知项；当训练一致时，其值趋于零；在插值情形下，等于超参数缩放的初始化噪声。有限样本伪证测试与种子扫查实验验证了该理论。
原文摘要 · Abstract (English)
The standard taxonomy of predictive uncertainty defines epistemic uncertainty as the part removable by collecting more data, while the standard measure identifies it with a mutual-information term. We prove the definition and the measure are extensionally inconsistent. On an explicit construction, the measure assigns all uncertainty to the epistemic class, yet no quantity of training data reduces it. Reducibility is instead a property of the pair (uncertainty, acquisition class), and the dichotomy resolves into three parts: aleatoric, sample-reducible epistemic, and mechanism-reducible epistemic uncertainty. An exact identity for the value of an observation shows that in-distribution data never reduces mechanism-irreducible uncertainty and generically increases it. Ensemble disagreement, the deployed epistemic estimate, tracks the training procedure rather than the epistemic term. It collapses to zero beneath a positive truth under consistent training, and equals hyperparameter-scaled initialization noise under interpolation. A finite-sample falsification test and seed-swept experiments confirm the theory.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。