Achieving Time Series Reasoning Requires Rethinking Model Design, Tasks Formulation, and Evaluation

Yaxuan Kong, Yiyuan Yang, Shiyu Wang, Chenghao Liu, Yuxuan Liang, Ming Jin, Stefan Zohren, Dan Pei, Yan Liu, Qingsong Wen · 2025

Read Achieving Time Series Reasoning Requires Rethinking Model Design, Tasks Formulation, and Evaluation by Yaxuan Kong, Yiyuan Yang, Shiyu Wang, Chenghao Liu, Yuxuan Liang, Ming Jin, Stefan Zohren, Dan Pei, Yan Liu, Qingsong Wen (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-01477.txt.

Description

Yaxuan Kong and nine co-authors argue time series AI needs rethinking in twenty twenty-five. Understanding time series is basic to many real applications. Multimodal language models have been tried as a way to add context beyond raw numbers. The field exploded from a handful of papers to hundreds in two years.

The authors analyze twenty influential works from twenty twenty-five. They find gaps in model design, task formulation, and evaluation alike. Methods borrow language techniques with little care for core time series traits. Tasks stay limited to prediction and classification. Evaluations prize benchmarks over robustness, interpretability, or decision value. Their call is to rethink all three together if true time series reasoning is the goal.

Catalog record

CREATORYaxuan Kong, Yiyuan Yang, Shiyu Wang, Chenghao Liu, Yuxuan Liang, Ming Jin, Stefan Zohren, Dan Pei, Yan Liu, Qingsong Wen
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE87 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-01477

Browse the full library · About the Tower of Babel