TS-FM Fine-Tuning Potential

Background: The adoption of general-purpose Time Series Foundation Models (TS-FMs) in domain-specific fields like transportation forecasting is a developing area.

Question / Future Work: Further gains in forecasting accuracy are anticipated when zero-shot TS-FMs, which currently show competitive or superior performance, are adapted using domain-specific data, prompting the need to investigate the efficacy and computational demands of fine-tuning techniques such as Parameter-Efficient Fine-Tuning (PEFT) in this context.