arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2609.15393cs.LGcs.AI

CodeTS:通过可执行代码实现可验证的文本到时间序列生成

CodeTS: Verifiable Text-to-Time Series Generation via Executable Code

Xudong Yuan, Shunyu Liu, Tongya Zheng, Huiping Zhuang, Mingli Song, Kaixuan Chen

首次发表
浏览论文内容

中文总结 AI 辅助

提出CodeTS框架,以代码为中间接口将文本到时间序列生成重构为文本-代码-时间序列流程,并设计多阶段执行奖励实现可验证的零样本生成,在八个基准上超越现有基线。

中文摘要 AI 辅助

文本到时间序列生成(Text-to-Time Series Generation, Text-to-TS)为从自然语言合成时间序列提供了一种有前景的范式,使得在真实观测数据稀缺或获取成本高昂时能够进行特定场景的生成。然而,现有方法通常缺乏从文本描述中推导生成逻辑以指导时间序列合成的显式机制。在本文中,我们提出了CodeTS,一个可验证的框架,它使用代码作为中间生成接口,将文本到时间序列生成重构为文本到代码再到时间序列(Text-to-Code-to-TS)的过程。CodeTS首先将文本时间描述映射到显式的代码空间,其中可执行代码指定了文本需求如何塑造目标时间模式,然后通过代码执行获得时间序列。为了在没有真实代码标注的情况下可靠地学习这一代码生成过程,CodeTS从结构化时间属性构建对齐的文本-代码-时间序列三元组,用于监督初始化。更重要的是,我们进一步设计了基于执行的多阶段奖励,用于验证格式有效性、代码可执行性和时间序列质量,从而使真实的文本-时间序列对能够为带可验证奖励的强化学习(RLVR)提供训练信号。在八个涵盖短、中、长生成长度的基准上的大量实验表明,CodeTS为文本到时间序列生成提供了强大的零样本解决方案,优于基于LLM的基线,并在目标数据集上训练的监督生成基线上取得了更好的平均结果。

英文摘要

Text-to-Time Series Generation (Text-to-TS) provides a promising paradigm for synthesizing time series from natural language, enabling scenario-specific generation when real observations are scarce or costly to acquire. However, existing methods typically lack an explicit mechanism for deriving generation logic from textual descriptions to guide time series synthesis. In this paper, we propose CodeTS, a verifiable framework that uses code as an intermediate generation interface, reformulating Text-to-TS generation as a Text-to-Code-to-TS process. CodeTS first maps textual temporal descriptions into an explicit code space, where executable code specifies how textual requirements shape target temporal patterns, and then obtains the time series through code execution. To learn this code generation process reliably without real code annotations, CodeTS constructs aligned Text-Code-TS triplets from structured temporal attributes for supervised initialization. More importantly, we further design multi-stage execution-based rewards that verify format validity, code executability, and time series quality, enabling real Text-TS pairs to provide training signals for Reinforcement Learning with Verifiable Rewards (RLVR). Extensive experiments on eight benchmarks across short, medium, and long generation lengths demonstrate that CodeTS provides a strong zero-shot solution for Text-to-TS generation, outperforming LLM-based baselines and achieving better averaged results than supervised generative baselines trained on the target datasets.

发表机构

  • Zhejiang University(浙江大学)
  • Nanyang Technological University(南洋理工大学)
  • South China University of Technology(华南理工大学)

机构由 AI 辅助整理,请以论文原文为准。

补充信息

↑