The paper introduces a framework called Faithful TimeSieve (FTS) to address the issue of unfaithfulness in the TimeSieve model, which is a state-of-the-art time series forecasting model. TimeSieve, while demonstrating impressive performance, exhibits instability due to sensitivity to random seeds, input perturbations, and parameter perturbations. The FTS framework aims to enhance the model's stability and faithfulness by defining three key attributes: Similarity in IB Space (Sib), Consistency in Prediction Space (Cps), and Stability in Noise Perturbations (Snp). These attributes ensure that the model retains essential information, maintains consistent predictions, and remains stable under perturbations. The framework is designed to integrate strategies that mitigate these issues while preserving the model's performance. Extensive experiments validate the effectiveness of the FTS framework, demonstrating improved faithfulness and robustness in TimeSieve's behavior. The results show that FTS significantly reduces the impact of random seeds, input perturbations, and parameter perturbations, leading to more reliable and trustworthy time series predictions.The paper introduces a framework called Faithful TimeSieve (FTS) to address the issue of unfaithfulness in the TimeSieve model, which is a state-of-the-art time series forecasting model. TimeSieve, while demonstrating impressive performance, exhibits instability due to sensitivity to random seeds, input perturbations, and parameter perturbations. The FTS framework aims to enhance the model's stability and faithfulness by defining three key attributes: Similarity in IB Space (Sib), Consistency in Prediction Space (Cps), and Stability in Noise Perturbations (Snp). These attributes ensure that the model retains essential information, maintains consistent predictions, and remains stable under perturbations. The framework is designed to integrate strategies that mitigate these issues while preserving the model's performance. Extensive experiments validate the effectiveness of the FTS framework, demonstrating improved faithfulness and robustness in TimeSieve's behavior. The results show that FTS significantly reduces the impact of random seeds, input perturbations, and parameter perturbations, leading to more reliable and trustworthy time series predictions.