Is More Context Always Better? Examining LLM Reasoning Capability for Time Interval Prediction
This work addresses the problem of temporal reasoning in LLMs for researchers and practitioners, revealing limitations in structured inference and guiding hybrid model design, though it is incremental in scope.
The study examined whether large language models (LLMs) can predict time intervals between recurring user actions, such as repurchases, and found that while LLMs outperform lightweight statistical baselines, they consistently underperform dedicated machine-learning models, with moderate context improving accuracy but excessive detail degrading it.
Large Language Models (LLMs) have demonstrated impressive capabilities in reasoning and prediction across different domains. Yet, their ability to infer temporal regularities from structured behavioral data remains underexplored. This paper presents a systematic study investigating whether LLMs can predict time intervals between recurring user actions, such as repeated purchases, and how different levels of contextual information shape their predictive behavior. Using a simple but representative repurchase scenario, we benchmark state-of-the-art LLMs in zero-shot settings against both statistical and machine-learning models. Two key findings emerge. First, while LLMs surpass lightweight statistical baselines, they consistently underperform dedicated machine-learning models, showing their limited ability to capture quantitative temporal structure. Second, although moderate context can improve LLM accuracy, adding further user-level detail degrades performance. These results challenge the assumption that "more context leads to better reasoning". Our study highlights fundamental limitations of today's LLMs in structured temporal inference and offers guidance for designing future context-aware hybrid models that integrate statistical precision with linguistic flexibility.