An Exploratory Study to Repurpose LLMs to a Unified Architecture for Time Series Classification
This work addresses the problem of improving time series classification for machine learning applications by identifying effective encoder choices, though it is incremental as it builds on prior repurposing efforts.
The study explored hybrid architectures combining specialized time series encoders with a frozen LLM backbone for time series classification, finding that only the Inception model consistently yielded positive performance gains.
Time series classification (TSC) is a core machine learning problem with broad applications. Recently there has been growing interest in repurposing large language models (LLMs) for TSC, motivated by their strong reasoning and generalization ability. Prior work has primarily focused on alignment strategies that explicitly map time series data into the textual domain; however, the choice of time series encoder architecture remains underexplored. In this work, we conduct an exploratory study of hybrid architectures that combine specialized time series encoders with a frozen LLM backbone. We evaluate a diverse set of encoder families, including Inception, convolutional, residual, transformer-based, and multilayer perceptron architectures, among which the Inception model is the only encoder architecture that consistently yields positive performance gains when integrated with an LLM backbone. Overall, this study highlights the impact of time series encoder choice in hybrid LLM architectures and points to Inception-based models as a promising direction for future LLM-driven time series learning.