Can LLMs Access their World Knowledge for Event Prediction?

Recent work has questioned whether large language models (LLMs) can be useful for temporal event prediction. Despite vast pretrained world knowledge, LLMs often underperform simpler specialized models.

近期的研究质疑大型语言模型（LLMs）在时间事件预测方面是否具有实用价值。尽管拥有海量的预训练世界知识，但 LLMs 的表现往往不如更简单的专用模型。

We ask whether this knowledge is truly inaccessible, or whether current methods fail to present event prediction in a form LLMs can use. We argue for the latter and show that LLMs become effective event forecasters when the interface is designed around two choices.

我们探讨了这些知识是真正无法获取，还是当前的方法未能以 LLMs 可用的形式呈现事件预测。我们支持后者，并证明当接口围绕两个选择进行设计时，LLMs 可以成为有效的事件预测器。

The first grounds each event in a natural-language description, and the second separates event semantics from the prediction space to prevent semantic bias from label associations. To test this, we introduce DescribeEvent, a benchmark of ~150K events across seven datasets and six domains with event-level descriptions.

第一个选择是将每个事件建立在自然语言描述的基础上，第二个选择是将事件语义与预测空间分离，以防止标签关联带来的语义偏差。为了验证这一点，我们引入了 DescribeEvent，这是一个包含七个数据集和六个领域中约 15 万个事件的基准测试，并提供了事件级别的描述。

We find that capable LLMs substantially benefit from descriptions and symbolic labels, outperforming description-free variants and non-LLM baselines, while weaker models gain little. These results suggest that accessing LLM world knowledge for event prediction is an interface problem, and that event descriptions offer a practical way to elicit the knowledge LLMs already encode.

我们发现，能力较强的 LLMs 能从描述和符号标签中获益良多，其表现优于无描述的变体和非 LLM 基准模型，而较弱的模型则获益甚微。这些结果表明，获取 LLM 的世界知识以进行事件预测是一个接口问题，而事件描述为激发 LLMs 已经编码的知识提供了一种实用的方法。