Lotus: Learning to Understand Time-series by Learning to Forecast

Computer Science > Machine Learning arXiv:2610.18877 (cs) [Submitted on 1 Oct 2026] Title: Lotus: Learning to Understand Time-series by Learning to Forecast Authors: Van Dai Do, Huu Hiep Nguyen, Minh Hoang Nguyen, Manh Nguyen, Hung Le.

计算机科学 > 机器学习 arXiv:2610.18877 (cs) [提交于 2026 年 10 月 1 日] 标题：Lotus：通过学习预测来学习理解时间序列 作者：Van Dai Do, Huu Hiep Nguyen, Minh Hoang Nguyen, Manh Nguyen, Hung Le。

Abstract: Time-series language models typically align numerical signals with language using labels, captions, or question-answer pairs. Although such supervision encourages linguistic representations, it does not explicitly require them to preserve information about how the underlying time-series signal evolves.

摘要：时间序列语言模型通常使用标签、标题或问答对将数值信号与语言对齐。尽管这种监督方式鼓励了语言表征，但它并未明确要求这些表征保留有关底层时间序列信号如何演变的信息。

We introduce forecast grounding, a pretraining principle that requires language-aligned time-series representations to remain predictive of the future evolution before they are adapted to downstream tasks. We instantiate this principle in LOTUS, which divides each time-series channel into non-overlapping patches and maps them into the LLM embedding space.

我们引入了“预测基础”（forecast grounding），这是一种预训练原则，要求与语言对齐的时间序列表征在适配下游任务之前，必须保持对未来演变的预测能力。我们在 LOTUS 中实例化了这一原则，它将每个时间序列通道划分为互不重叠的补丁（patches），并将它们映射到大语言模型（LLM）的嵌入空间中。

It then predicts the next patch from the LLM hidden states of the time-series tokens, encouraging the learned representation to retain temporal information. To capture both temporal changes and periodic patterns, we also propose a temporal-spectral patch representation that combines each raw time-series patch with its spectral representation before mapping it into the LLM embedding space.

随后，它通过时间序列标记（tokens）的 LLM 隐藏状态来预测下一个补丁，从而鼓励学习到的表征保留时间信息。为了同时捕捉时间变化和周期性模式，我们还提出了一种时频补丁表征，在将其映射到 LLM 嵌入空间之前，将每个原始时间序列补丁与其频谱表征相结合。

Across three language model backbones and six downstream evaluation datasets, LOTUS achieves the best average performance with minimal trainable parameters, reaching 59.91% accuracy and outperforming the strongest matched baseline by 3.36%. Our representation analyses further show that LOTUS uses the numerical input and preserves temporal structure in its learned patch representations.

在三个语言模型主干和六个下游评估数据集上，LOTUS 以最少的可训练参数实现了最佳的平均性能，准确率达到 59.91%，比最强的匹配基准高出 3.36%。我们的表征分析进一步表明，LOTUS 使用了数值输入，并在其学习到的补丁表征中保留了时间结构。

These results establish forecast grounding as a promising pretraining principle, enabling compact time-series language models to outperform models 25 to 50 times their size on TSRBench.

这些结果确立了“预测基础”作为一种有前景的预训练原则，使紧凑型时间序列语言模型能够在 TSRBench 上超越规模为其 25 到 50 倍的模型。