Can LLMs Access their World Knowledge for Event Prediction?
Author1, Author2, Author3, Author4, Author5
Abstract
The paper investigates the effectiveness of large language models (LLMs) in temporal event prediction and suggests that their underperformance is due to interface design rather than a lack of world knowledge.
Reality Card
LLMs can become effective event forecasters when provided with natural-language descriptions and a separation of event semantics from prediction space.
LLMs substantially outperform description-free variants and non-LLM baselines when using event descriptions, with a benchmark of ~150K events tested.
The study may be limited by the specific design of the DescribeEvent benchmark and the potential variability in LLM capabilities.
Paper to code
Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.