Cisco Time Series Model Technical Report

2025-11-24

AI Summary

This work addresses the accuracy and generalization bottlenecks in zero-shot long-horizon univariate time series forecasting—particularly under multi-granularity observability scenarios. We propose the first univariate foundation model explicitly designed for zero-shot cross-resolution forecasting. Methodologically, we introduce a decoder-only multi-resolution architecture capable of handling heterogeneous temporal granularities, built upon an enhanced TimesFM framework and pretrained via large-scale self-supervision on over 300 billion time points. Our key innovation is a resolution-aware contextual modeling mechanism that jointly enables long-range dependency capture and zero-shot cross-granularity transfer. Experiments demonstrate significant improvements over state-of-the-art methods on the Observability dataset, while maintaining competitive performance on the general-purpose GIFT-Eval benchmark—validating the model’s strong generalization capability and domain adaptability.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Computer Vision: Other Foundations of Computer Vision
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

We introduce the Cisco Time Series Model, a univariate zero-shot forecaster. This time series foundation model is the result of a general architectural innovation to a time series model enabling it to accept multiresolution input, applied to a popular decoder-only time series model (TimesFM). The resulting multiresolution decoder-only model is trained on over 300B unique data points, with more than half coming from the observability domain. Quantitative and qualitative evaluations demonstrate that the resulting model achieves superior performance on observability datasets while retaining very similar performance on a standard general-purpose forecasting benchmark (GIFT-Eval), and suggest that the multiresolution structure enables the model to make more accurate predictions on long context input.

Problem

• Develops a zero-shot univariate time series forecasting model
• Enables multiresolution input processing for time series data
• Improves prediction accuracy on long context observability datasets

Innovation

• Univariate zero-shot forecasting foundation model
• Multiresolution decoder-only architecture innovation
• Trained on 300B unique observability data points