Understanding Why Large Language Models Can Be Ineffective in Time Series Analysis: The Impact of Modality Alignment
This work addresses the problem of inefficient model selection in time series analysis for researchers and practitioners, showing that LLMs are often ineffective compared to simpler alternatives, making it an incremental contribution by challenging current trends.
The study found that large language models (LLMs) offer minimal advantages for time series tasks like forecasting and classification, with simpler models consistently outperforming them while using far fewer parameters, and revealed that reprogramming techniques fail to effectively align time series data with language, showing 'pseudo-alignment' behavior.
Large Language Models (LLMs) have demonstrated impressive performance in time series analysis and seems to understand the time temporal relationship well than traditional transformer-based approaches. However, since LLMs are not designed for time series tasks, simpler models like linear regressions can often achieve comparable performance with far less complexity. In this study, we perform extensive experiments to assess the effectiveness of applying LLMs to key time series tasks, including forecasting, classification, imputation, and anomaly detection. We compare the performance of LLMs against simpler baseline models, such as single layer linear models and randomly initialized LLMs. Our results reveal that LLMs offer minimal advantages for these core time series tasks and may even distort the temporal structure of the data. In contrast, simpler models consistently outperform LLMs while requiring far fewer parameters. Furthermore, we analyze existing reprogramming techniques and show, through data manifold analysis, that these methods fail to effectively align time series data with language and display "pseudo-alignment" behavior in embedding space. Our findings suggest that the performance of LLM based methods in time series tasks arises from the intrinsic characteristics and structure of time series data, rather than any meaningful alignment with the language model architecture.