A decoder-only foundation model for time-series forecasting

Papers with Code Trending Papers

Summary

This article presents a research paper on Time-Series Foundation Model (TimeFM), a decoder-only model that achieves near-optimal zero-shot performance across diverse time-series datasets by adapting large language model techniques.

Motivated by recent advances in large language models for Natural Language Processing (NLP), we design a time-series foundation model for forecasting whose out-of-the-box zero-shot performance on a variety of public datasets comes close to the accuracy of state-of-the-art supervised forecasting models for each individual dataset. Our model is based on pretraining a patched-decoder style attention model on a large time-series corpus, and can work well across different forecasting history lengths, prediction lengths and temporal granularities.
Original Article
View Cached Full Text

Cached at: 05/08/26, 08:47 AM

Paper page - A decoder-only foundation model for time-series forecasting

Source: https://huggingface.co/papers/2310.10688 Published on Oct 14, 2023

Abstract

A large language model adapted for time-series forecasting achieves near-optimal zero-shot performance on diverse datasets across different time scales and granularities.

Motivated by recent advances in large language models for Natural Language Processing (NLP), we design a time-series foundation model forforecastingwhose out-of-the-boxzero-shot performanceon a variety of public datasets comes close to the accuracy of state-of-the-art supervisedforecastingmodels for each individual dataset. Our model is based on pretraining apatched-decoderstyleattention modelon a largetime-series corpus, and can work well across differentforecastinghistory lengths, prediction lengths and temporal granularities.

View arXiv pageView PDFGitHub19.5kautoAdd to collection

Community

Upload images, audio, and videos by dragging in the text input, pasting, orclicking here.

Tap or paste here to upload images

Get this paper in your agent:

hf papers read 2310\.10688

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper14

#### google/timesfm-1.0-200m Time Series Forecasting• UpdatedMay 17, 2024 • 1.3k • 814 #### google/timesfm-2.0-500m-pytorch Time Series Forecasting• 0.5B• UpdatedApr 16, 2025 • 35.7k • 252 #### google/timesfm-2.5-200m-pytorch Time Series Forecasting• UpdatedOct 2, 2025 • 228 #### google/timesfm-2.5-200m-transformers Time Series Forecasting• 0.2B• Updated27 days ago • 145k • 82 Browse 14 models citing this paper## Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2310.10688 in a dataset README.md to link it from this page.

Spaces citing this paper46

Collections including this paper9

Browse 9 collections that include this paper

Similar Articles

Unified Zero-Shot Time Series Forecasting: A Darts Foundation

arXiv cs.LG

Darts, a popular open-source Python library for time series analysis, introduces a unified FoundationModel class collection that integrates multiple time series foundation models (Chronos-2, TimesFM 2.5, TiRex, PatchTST-FM) for zero-shot and fine-tuned forecasting with standardized interfaces and minimal dependencies.