In long-term time series forecasting, different variables often influence the target variable over distinct time intervals, a challenge known as the multi-delay issue. Traditional models typically process all variables or time points uniformly, which limits their ability to capture complex variable relationships and obtain non-trivial time representations. To address this issue, we propose TimePro, an innovative Mamba-based model that constructs variate- and time-aware hyper-states. Unlike conventional approaches that merely transfer plain states across variable or time dimensions, TimePro preserves the fine-grained temporal features of each variate token and adaptively selects the focused time points to tune the plain state. The reconstructed hyper-state can perceive both variable relationships and salient temporal information, which helps the model make accurate forecasting. In experiments, TimePro performs competitively on eight real-world long-term forecasting benchmarks with satisfactory linear complexity. Code is available at https://github.com/xwmaxwma/TimePro.

In long-term time series forecasting, different variables often influence the target variable over distinct time intervals, a challenge known as the multi-delay issue. We design the model for the multi-delay issue thus improving the prediction performance.We propose the TimePro from the most recent Mamba. Specifically, we design a variable- and time-aware hyper-state that senses both variable relationships and significant temporal information within variables to effectively solve the multi-latency issue.Our model TimePro can significantly improve the performance of multivariate time series forecasting. In particular, it also possesses good efficiency. We have also verified that TimePro can be used as a powerful and efficient predictor in some real-world industrial scenarios .