两篇新的arXiv论文探讨了机器学习中顺序决策的理论框架。第一篇论文引入了一个“机制信息”度量，用于量化结合了物理先验和学习残差的混合模型的价值，在模拟中展示了样本效率的提高，并警告不要在安全关键型应用中使用LLM先验。第二篇论文开发了一个顺序超采样框架，为自适应数据设置建立信息论泛化界限，适用于在线学习、流式主动学习和多臂老虎机。 AI
影响 这些论文在理解和界定顺序决策模型性能方面提供了理论进展，可能影响未来在数据稀缺或安全关键领域AI系统的设计。
排序理由 两篇发表在arXiv上的学术论文，提出了新的顺序决策理论框架。
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
Hybrid mechanistic models, physical priors with learned residuals, promise to reduce the data required for good decisions, but have no computable criterion to test this. We characterize the value of mechanistic priors in sequential decision-making within both asymptotic and burn-…
arXiv:2605.12190v1 Announce Type: new Abstract: Information-theoretic generalization bounds based on the supersample construction are a central tool for algorithm-dependent generalization analysis in the batch i.i.d.~setting. However, existing supersample conditional mutual infor…
Information-theoretic generalization bounds based on the supersample construction are a central tool for algorithm-dependent generalization analysis in the batch i.i.d.~setting. However, existing supersample conditional mutual information (CMI) bounds do not directly apply to seq…