标题： 不确定性驱动的可靠性：现代机器学习中的选择性预测与可信部署 显示英文标题

标题： Uncertainty-Driven Reliability: Selective Prediction and Trustworthy Deployment in Modern Machine Learning

摘要： 机器学习（ML）系统在对可靠性要求很高的领域中得到越来越多的部署。 本论文研究了不确定性估计如何增强ML的安全性和可信度，重点在于选择性预测——当置信度低时模型会回避预测。 我们首先表明，模型的训练轨迹包含丰富的不确定性信号，可以在不改变其架构或损失函数的情况下加以利用。 通过集成中间检查点的预测，我们提出了一种轻量级、后期的回避方法，该方法适用于各种任务，避免了深度集成的成本，并实现了最先进的选择性预测性能。 关键的是，这种方法与差分隐私（DP）完全兼容，使我们能够研究隐私噪声如何影响不确定性质量。 我们发现，尽管许多方法在DP下性能下降，但我们的基于轨迹的方法依然稳健，并我们引入了一个框架来隔离隐私-不确定性权衡。 接下来，我们开发了选择性分类差距的有限样本分解——即偏离理想准确率-覆盖率曲线的偏差——识别出五个可解释的误差来源，并阐明哪些干预措施可以缩小差距。 这解释了为什么仅校准无法修复排序错误，从而促使改进不确定性排序的方法。 最后，我们展示了不确定性信号可以被对抗性操纵以隐藏错误或拒绝服务，同时保持高准确性，并我们设计了结合校准审计和可验证推理的防御措施。 总的来说，这些贡献通过改进、评估和保护不确定性估计来推动可靠的ML，使模型不仅能够做出准确的预测——而且知道何时说“我不知道”。 显示英文摘要

摘要： Machine learning (ML) systems are increasingly deployed in high-stakes domains where reliability is paramount. This thesis investigates how uncertainty estimation can enhance the safety and trustworthiness of ML, focusing on selective prediction -- where models abstain when confidence is low. We first show that a model's training trajectory contains rich uncertainty signals that can be exploited without altering its architecture or loss. By ensembling predictions from intermediate checkpoints, we propose a lightweight, post-hoc abstention method that works across tasks, avoids the cost of deep ensembles, and achieves state-of-the-art selective prediction performance. Crucially, this approach is fully compatible with differential privacy (DP), allowing us to study how privacy noise affects uncertainty quality. We find that while many methods degrade under DP, our trajectory-based approach remains robust, and we introduce a framework for isolating the privacy-uncertainty trade-off. Next, we then develop a finite-sample decomposition of the selective classification gap -- the deviation from the oracle accuracy-coverage curve -- identifying five interpretable error sources and clarifying which interventions can close the gap. This explains why calibration alone cannot fix ranking errors, motivating methods that improve uncertainty ordering. Finally, we show that uncertainty signals can be adversarially manipulated to hide errors or deny service while maintaining high accuracy, and we design defenses combining calibration audits with verifiable inference. Together, these contributions advance reliable ML by improving, evaluating, and safeguarding uncertainty estimation, enabling models that not only make accurate predictions -- but also know when to say "I do not know".