arXiv：cs.CL· Zijun Wu, Zhitian Zhang, Hossein Hajimirsadeghi, Nikita Dvornik, Alexander Pashevich·· 16 小时前
LLM 能否调用世界知识做事件预测？DescribeEvent 基准与接口设计给出答案
Can LLMs Access their World Knowledge for Event Prediction?
AI 导读
研究提出 DescribeEvent 基准，包含约 150K 事件、覆盖七个数据集和六个领域，用于检验 LLM 的事件预测能力。结果显示，将事件用自然语言描述表达、并把事件语义与预测空间分离后，能力较强的 LLM 显著优于无描述变体和非 LLM 基线，较弱模型则提升有限。作者据此认为，LLM 事件预测表现不佳是接口设计问题，而非世界知识不可用。
正文
Abstract: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. 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. 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. 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.
| Comments: | Accepted to Findings of EMNLP 2026 |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.19021 [cs.CL] |
| (or arXiv:2610.19021v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.19021 arXiv-issued DOI via DataCite |
Submission history
From: Zijun Wu [view email]
[v1]
Tue, 13 Oct 2026 18:14:05 UTC (281 KB)
来源：arXiv：cs.CL · arxiv.org