UniFlow: Unified Multi-Scale Flow Matching for Time Series Generation
Abstract
Real-world time series exhibit complex dynamics across multiple temporal resolutions, yet effectively incorporating multi-scale information into generative modeling remains challenging without increasing generative complexity. We propose UniFlow, a unified framework that integrates multi-scale representation learning with efficient non-autoregressive generation through a shared-semantic discrete interface. Specifically, a multi-scale tokenizer employs shared-index voting to consolidate complementary temporal information into a single public-ID sequence, while scale-specific codebooks preserve the information required for reconstruction. Building on this unified representation, we develop geometry-aware categorical flow matching, which initializes transport from learned low-rank anchors and generates public identities non-autoregressively. This design captures multi-scale temporal structure without requiring multiple generative token streams or sequential token decoding. Experiments across four benchmarks and both short- and long-term settings demonstrate improved generative fidelity, strong cross-scale consistency, and substantial inference speedups over representative autoregressive and diffusion-based baselines. Code is available at https://github.com/William-Liwei/uniflow
Related
- A Unified Framework for Assortment Optimization and Matching under MNL Choice
- A Unified Dual Method for Matching Problems
- Generalized Wasserstein Flow Matching: Nested Couplings for Point Cloud Generation
- Counterfactual Generation via Flow Matching: Coupling-Sensitive End-to-End Rates
- Gradient Discrete Schrodinger Bridge Matching for Peptide Generation
BibTeX
@misc{feng2026uniflow,
title = {{UniFlow: Unified Multi-Scale Flow Matching for Time Series Generation}},
author = {Shibo Feng and Wei Li and Min Wu and Chunyan Miao and Peilin Zhao},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.04825},
eprint = {2611.04825},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.04825},
}
RIS
TY - UNPB
TI - UniFlow: Unified Multi-Scale Flow Matching for Time Series Generation
AU - Shibo Feng
AU - Wei Li
AU - Min Wu
AU - Chunyan Miao
AU - Peilin Zhao
PY - 2026
DA - 2026/11/03
PB - arXiv
DO - 10.48550/arxiv.2611.04825
UR - https://arxiv.org/abs/2611.04825
AB - Real-world time series exhibit complex dynamics across multiple temporal resolutions, yet effectively incorporating multi-scale information into generative modeling remains challenging without increasing generative complexity. We propose UniFlow, a unified framework that integrates multi-scale representation learning with efficient non-autoregressive generation through a shared-semantic discrete interface. Specifically, a multi-scale tokenizer employs shared-index voting to consolidate complementary temporal information into a single public-ID sequence, while scale-specific codebooks preserve the information required for reconstruction. Building on this unified representation, we develop geometry-aware categorical flow matching, which initializes transport from learned low-rank anchors and generates public identities non-autoregressively. This design captures multi-scale temporal structure without requiring multiple generative token streams or sequential token decoding. Experiments across four benchmarks and both short- and long-term settings demonstrate improved generative fidelity, strong cross-scale consistency, and substantial inference speedups over representative autoregressive and diffusion-based baselines. Code is available at https://github.com/William-Liwei/uniflow
ER -