When Denoising Hurts: Rethinking the Terminal Step of Diffusion Time Series Forecasters -- Extended Version

2026-08-14

AI Summary

This study addresses the statistical drift and performance degradation caused by persistent low-noise denoising in diffusion-based time series forecasting by elucidating the detrimental mechanisms of over-denoising. We propose a label-free global stopping criterion and a Bernoulli time-step sampler focusing on high-noise regions to jointly optimize training sampling distributions and inference termination points. Experiments across eight real-world datasets demonstrate that this approach effectively circumvents the over-denoising trap, significantly improving prediction accuracy while accelerating inference. The proposed method achieves superior overall performance compared to existing mainstream techniques, establishing a new paradigm for the efficient application of diffusion models in time series forecasting.

Abstract

Diffusion models offer a natural way to model uncertainty in time series forecasting, yet their iterative sampling process is often treated as a uniformly beneficial refinement procedure. Our study challenges this view by examining how forecast quality evolves throughout reverse diffusion. We find that general temporal structure is often recovered at relatively high noise levels, whereas continued low-noise refinement can introduce statistical drift and degrade the final forecast. Our analysis further suggests that this behavior explains why prior methods often favor relatively narrow diffusion architecture and schedule design. Building on this observation, we propose a label-free global stopping criterion that detects the optimal termination point, eventually speeding up inference and improving predictive accuracy. Additionally, since early stopping terminates inference in high-noise regions, we propose a Bernoulli timestep sampler that concentrates training on this region while preserving coverage of the full diffusion process. Extensive experiments conducted across eight real-world datasets demonstrate the superior performance of our method compared to existing approaches.

Problem

• Diffusion Models
• Time Series Forecasting
• Statistical Drift
• Reverse Diffusion
• Optimal Stopping

Innovation

• Label-free Global Stopping Criterion
• Bernoulli Timestep Sampler
• Statistical Drift
• Diffusion Time Series Forecasting