包裹，原文开头的第一部分是一个完整的自然段落，所以用
包起来。然后接下来的“Key contributions”后面的内容是列表，所以需要用
-
和
- 来处理，每个贡献点作为
- 里的内容。接下来的“Pre-training results”部分是另一个自然段落吗？不对，看原文，“We find that pre-trained uncertainty estimators...”这部分是一个自然段落吗？不，原文的结构是：第一个段落之后，有两个贡献点的列表，然后下一段是“We pre-train several UQ models... We find that pre-trained uncertainty estimators...”这一整段是第二个自然段落吗？等一下，仔细看原文： 原文的结构是： 1. 第一个段落：Informative uncertainty quantification (UQ)... narrow domains of interest. Inspired by... broader applicability. 然后是Key contributions: - We pre-train several UQ models on large materials science datasets to quantify the uncertainty of different MLIPs, and evaluate their performance on both in-distribution and out-of-domain (OOD) adsorption-energy predictions. - We pre-train several UQ models to serve as foundational UQ models for MLIPs, enabling zero-shot uncertainty estimates and providing a priori information about model reliability and applicability. 接下来是：We find that pre-trained uncertainty estimators... These results lay the groundwork... practical settings. 然后还有一段：We pre-train several UQ models on large materials science datasets to quantify the uncertainty of different MLIPs, and evaluate their performance on both in-distribution and out-of-domain (OOD) adsorption-energy predictions. We find that pre-trained uncertainty estimators provide viable zero-shot uncertainty estimates across multiple datasets, while fine-tuning consistently yields positive correlations between predicted uncertainty and prediction error, even on the most challenging datasets. 哦，可能原文的结构是： - 第一段：讲背景、现有方法局限、提出的预训练思路。 - 然后是Key contributions列表，两个点。 - 然后是第二段：讲预训练的数据集、评估方式，以及发现。 - 然后是第三段：总结结果的意义。 不对，再仔细看用户给的原文： 用户给的原文是： Informative uncertainty quantification (UQ) is essential for the practical deployment of machine-learned interatomic potentials (MLIPs), as it provides insight into their reliability across chemical space. Although numerous post-hoc uncertainty estimation methods for MLIPs have recently been proposed, most have been trained in low-data regimes, restricting their applicability to narrow domains of interest. Inspired by the emergence of foundational MLIPs, we propose pre-training uncertainty estimators on large, diverse datasets to enable broader applicability. We pre-train several UQ models on large materials science datasets to quantify the uncertainty of different MLIPs, and evaluate their performance on both in-distribution and out-of-domain (OOD) adsorption-energy predictions. We find that pre-trained uncertainty estimators provide viable zero-shot uncertainty estimates across multiple datasets, while fine-tuning consistently yields positive correlations between predicted uncertainty and prediction error, even on the most challenging datasets. These results lay the groundwork for foundational UQ models for MLIPs, ultimately facilitating their deployment in practical settings by providing users with a priori information about model reliability and applicability. 哦，原来如此，可能用户给的原文里的换行是段落分隔？但用户给的原文里的换行可能是输入时的换行，实际应该是几个段落？不对，看用户给的原文的格式： 用户给的原文是： Informative uncertainty quantification (UQ) is essential for the practical deployment of machine-learned interatomic potentials (MLIPs), as it provides insight into their reliability across chemical space. Although numerous post-hoc uncertainty estimation methods for MLIPs have recently been proposed, most have been trained in low-data regimes, restricting their applicability to narrow domains of interest. Inspired by the emergence of foundational MLIPs, we propose pre-training uncertainty estimators on large, diverse datasets to enable broader applicability. We pre-train several UQ models on large materials science datasets to quantify the uncertainty of different MLIPs, and evaluate their performance on both in-distribution and out-of-domain (OOD) adsorption-energy predictions. We find that pre-trained uncertainty estimators provide viable zero-shot uncertainty estimates across multiple datasets, while fine-tuning consistently yields positive correlations between predicted uncertainty and prediction error, even on the most challenging datasets. These results lay the groundwork for foundational UQ models for MLIPs, ultimately facilitating their deployment in practical settings by providing users with a priori information about model reliability and applicability. 然后接下来是： Key contributions: - We pre-train several UQ models on large materials science datasets to quantify the uncertainty of different MLIPs, and evaluate their performance on both in-distribution and out-of-domain (OOD) adsorption-energy predictions. - We pre-train several UQ models to serve as foundational UQ models for MLIPs, enabling zero-shot uncertainty estimates and providing a priori information about model reliability and applicability. Pre-training results: - We find that pre-trained uncertainty estimators provide viable zero-shot uncertainty estimates across multiple datasets, while fine-tuning consistently yields positive correlations between predicted uncertainty and prediction error, even on the most challenging datasets. - These results lay the groundwork for foundational UQ models for MLIPs, ultimately facilitating their deployment in practical settings by providing users with a priori information about model reliability and applicability. 哦，原来用户给的原文里，前面有一个段落，然后是Key contributions的列表，然后是Pre-training results的列表？不对，再看用户给的原文的完整内容： 用户的原文是： Informative uncertainty quantification (UQ) is essential for the practical deployment of machine-learned interatomic potentials (MLIPs), as it provides insight into their reliability across chemical space. Although numerous post-hoc uncertainty estimation methods for MLIPs have recently been proposed, most have been trained in low-data regimes, restricting their applicability to narrow domains of interest. Inspired by the emergence of foundational MLIPs, we propose pre-training uncertainty estimators on large, diverse datasets to enable broader applicability. We pre-train several UQ models on large materials science datasets to quantify the uncertainty of different MLIPs, and evaluate their performance on both in-distribution and out-of-domain (OOD) adsorption-energy predictions. We find that pre-trained uncertainty estimators provide viable zero-shot uncertainty estimates across multiple datasets, while fine-tuning consistently yields positive correlations between predicted uncertainty and prediction error, even on the most challenging datasets. These results lay the groundwork for foundational UQ models for MLIPs, ultimately facilitating their deployment in practical settings by providing users with a priori information about model reliability and applicability. Key contributions: - We pre-train several UQ models on large materials science datasets to quantify the uncertainty of different MLIPs, and evaluate their performance on both in-distribution and out-of-domain (OOD) adsorption-energy predictions. - We pre-train several UQ models to serve as foundational UQ models for MLIPs, enabling zero-shot uncertainty estimates and providing a priori information about model reliability and applicability. Pre-training results: - We find that pre-trained uncertainty estimators provide viable zero-shot uncertainty estimates across multiple datasets, while fine-tuning consistently yields positive correlations between predicted uncertainty and prediction error, even on the most challenging datasets. - These results lay the groundwork for foundational UQ models for MLIPs, ultimately facilitating their deployment in practical settings by providing users with a priori information about model reliability and applicability. 哦，原来用户给的原文是这样的结构：前面三个段落？不，可能用户的原文里，前面三个部分是段落，然后是Key contributions和Pre-training results的列表。但根据规则，长文用
/
划分小节，所以Key contributions和Pre-training results应该作为小节标题，用
或者
？规则里说“长文用
/
划分小节”，所以这里的Key contributions和Pre-training results可以作为
的标题。 然后，第一个自然段落是：“Informative uncertainty quantification (UQ)... broader applicability.” 这部分