Saved in:
Bibliographic Details
Main Authors: Li, Siyuan, Sun, Yifan, Cheng, Lei, Wang, Lewen, Liu, Yang, Liu, Weiqing, Li, Jianlong, Bian, Jiang, Fang, Shikai
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2511.04973
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911253445738496
author Li, Siyuan
Sun, Yifan
Cheng, Lei
Wang, Lewen
Liu, Yang
Liu, Weiqing
Li, Jianlong
Bian, Jiang
Fang, Shikai
author_facet Li, Siyuan
Sun, Yifan
Cheng, Lei
Wang, Lewen
Liu, Yang
Liu, Weiqing
Li, Jianlong
Bian, Jiang
Fang, Shikai
contents Generative models for multivariate time series are essential for data augmentation, simulation, and privacy preservation, yet current state-of-the-art diffusion-based approaches are slow and limited to fixed-length windows. We propose FAR-TS, a simple yet effective framework that combines disentangled factorization with an autoregressive Transformer over a discrete, quantized latent space to generate time series. Each time series is decomposed into a data-adaptive basis that captures static cross-channel correlations and temporal coefficients that are vector-quantized into discrete tokens. A LLaMA-style autoregressive Transformer then models these token sequences, enabling fast and controllable generation of sequences with arbitrary length. Owing to its streamlined design, FAR-TS achieves orders-of-magnitude faster generation than Diffusion-TS while preserving cross-channel correlations and an interpretable latent space, enabling high-quality and flexible time series synthesis.
format Preprint
id arxiv_https___arxiv_org_abs_2511_04973
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Less Is More: Generating Time Series with LLaMA-Style Autoregression in Simple Factorized Latent Spaces
Li, Siyuan
Sun, Yifan
Cheng, Lei
Wang, Lewen
Liu, Yang
Liu, Weiqing
Li, Jianlong
Bian, Jiang
Fang, Shikai
Machine Learning
Generative models for multivariate time series are essential for data augmentation, simulation, and privacy preservation, yet current state-of-the-art diffusion-based approaches are slow and limited to fixed-length windows. We propose FAR-TS, a simple yet effective framework that combines disentangled factorization with an autoregressive Transformer over a discrete, quantized latent space to generate time series. Each time series is decomposed into a data-adaptive basis that captures static cross-channel correlations and temporal coefficients that are vector-quantized into discrete tokens. A LLaMA-style autoregressive Transformer then models these token sequences, enabling fast and controllable generation of sequences with arbitrary length. Owing to its streamlined design, FAR-TS achieves orders-of-magnitude faster generation than Diffusion-TS while preserving cross-channel correlations and an interpretable latent space, enabling high-quality and flexible time series synthesis.
title Less Is More: Generating Time Series with LLaMA-Style Autoregression in Simple Factorized Latent Spaces
topic Machine Learning
url https://arxiv.org/abs/2511.04973