Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

TF-CoDiT: Conditional Time Series Synthesis with Diffusion Transformers for Treasury Futures

Yingxiao Zhang, Jiaxin Duan, Junfu Zhang, Ke Feng

TL;DR

This work tackles the challenge of generating realistic treasury futures time series under sparse data and exogenous macro conditions. It introduces TF-CoDiT, a text-to-time-series diffusion framework that combines Discrete Wavelet Transform representations, a U-shape VAE for cross-channel encoding, and a backbone LLM guided by FinMAP prompts to produce conditional, high-fidelity TF trajectories. Empirical results show state-of-the-art fidelity, with mean errors reaching as low as $\text{MSE} = 0.433$ and $\text{MAE} = 0.453$, and substantial improvements over baselines like TimeLLM and T2S, particularly over longer horizons. The approach provides a practical, linguistically grounded mechanism for controllable synthesis of multi-variable treasury futures, with demonstrated robustness across contracts and time scales and potential for broader financial-domain conditioning.

Abstract

Diffusion Transformers (DiT) have achieved milestones in synthesizing financial time-series data, such as stock prices and order flows. However, their performance in synthesizing treasury futures data is still underexplored. This work emphasizes the characteristics of treasury futures data, including its low volume, market dependencies, and the grouped correlations among multivariables. To overcome these challenges, we propose TF-CoDiT, the first DiT framework for language-controlled treasury futures synthesis. To facilitate low-data learning, TF-CoDiT adapts the standard DiT by transforming multi-channel 1-D time series into Discrete Wavelet Transform (DWT) coefficient matrices. A U-shape VAE is proposed to encode cross-channel dependencies hierarchically into a latent variable and bridge the latent and DWT spaces through decoding, thereby enabling latent diffusion generation. To derive prompts that cover essential conditions, we introduce the Financial Market Attribute Protocol (FinMAP) - a multi-level description system that standardizes daily$/$periodical market dynamics by recognizing 17$/$23 economic indicators from 7/8 perspectives. In our experiments, we gather four types of treasury futures data covering the period from 2015 to 2025, and define data synthesis tasks with durations ranging from one week to four months. Extensive evaluations demonstrate that TF-CoDiT can produce highly authentic data with errors at most 0.433 (MSE) and 0.453 (MAE) to the ground-truth. Further studies evidence the robustness of TF-CoDiT across contracts and temporal horizons.

TF-CoDiT: Conditional Time Series Synthesis with Diffusion Transformers for Treasury Futures

TL;DR

This work tackles the challenge of generating realistic treasury futures time series under sparse data and exogenous macro conditions. It introduces TF-CoDiT, a text-to-time-series diffusion framework that combines Discrete Wavelet Transform representations, a U-shape VAE for cross-channel encoding, and a backbone LLM guided by FinMAP prompts to produce conditional, high-fidelity TF trajectories. Empirical results show state-of-the-art fidelity, with mean errors reaching as low as and , and substantial improvements over baselines like TimeLLM and T2S, particularly over longer horizons. The approach provides a practical, linguistically grounded mechanism for controllable synthesis of multi-variable treasury futures, with demonstrated robustness across contracts and time scales and potential for broader financial-domain conditioning.

Abstract

Diffusion Transformers (DiT) have achieved milestones in synthesizing financial time-series data, such as stock prices and order flows. However, their performance in synthesizing treasury futures data is still underexplored. This work emphasizes the characteristics of treasury futures data, including its low volume, market dependencies, and the grouped correlations among multivariables. To overcome these challenges, we propose TF-CoDiT, the first DiT framework for language-controlled treasury futures synthesis. To facilitate low-data learning, TF-CoDiT adapts the standard DiT by transforming multi-channel 1-D time series into Discrete Wavelet Transform (DWT) coefficient matrices. A U-shape VAE is proposed to encode cross-channel dependencies hierarchically into a latent variable and bridge the latent and DWT spaces through decoding, thereby enabling latent diffusion generation. To derive prompts that cover essential conditions, we introduce the Financial Market Attribute Protocol (FinMAP) - a multi-level description system that standardizes dailyperiodical market dynamics by recognizing 1723 economic indicators from 7/8 perspectives. In our experiments, we gather four types of treasury futures data covering the period from 2015 to 2025, and define data synthesis tasks with durations ranging from one week to four months. Extensive evaluations demonstrate that TF-CoDiT can produce highly authentic data with errors at most 0.433 (MSE) and 0.453 (MAE) to the ground-truth. Further studies evidence the robustness of TF-CoDiT across contracts and temporal horizons.
Paper Structure (32 sections, 21 equations, 7 figures, 14 tables)

This paper contains 32 sections, 21 equations, 7 figures, 14 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Diffusion Transformers
  4. Financial Times-series Synthesis
  5. Methodology
  6. Problem Definition
  7. Signal Transformation
  8. Backbone LLM
  9. U-shape Variational Autoencoder
  10. FinMAP
  11. Inference
  12. Experiment
  13. Datasets
  14. Experimental Settings
  15. Evaluation Results
  16. ...and 17 more sections

Figures (7)

  • Figure 1: Monthly tendency line of treasury futures in four types of contracts (T, TS, TF, and TL).
  • Figure 2: Overview of the TF-CoDiT framework.
  • Figure 3: Architecture of FinMAP taxonomy.
  • Figure 4: Error band analyses in the contract-T.
  • Figure 5: Element-wise reconstruction errors achieved by U-VAE with MSE loss (upper) and L1 loss (lower).
  • ...and 2 more figures