ABSTRACT
This research paper delves into the types of errors in temporal reasoning models, categorizing them into causal inversion, temporal anachronism, and dependency violation. It presents a comparative analysis of different models and datasets, highlighting the number of errors and the breakdown by type.
PAPER · PDF
Loading PDF...
Key findings
Identified three main error types in temporal reasoning: causal inversion, temporal anachronism, and dependency violation.
Presented a comparative analysis of error rates across different models and datasets.
Highlighted the importance of understanding error types for improving model accuracy in temporal reasoning tasks.
Limitations & open questions
Limited to specific models and datasets, broader analysis needed.
Further research required to mitigate identified error types.