Empowering VLMs for Few-Shot Multimodal Time Series Classification via Tailored Agentic Reasoning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Li, Lin, Huang, Jiawei, Quan, Qihao, Li, Dan, Li, Boxin, Zhang, Xiao, Meng, Erli, Feng, Wenjie, Lou, Jian, Ng, See-Kiong
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910224419389440
author Li, Lin
Huang, Jiawei
Quan, Qihao
Li, Dan
Li, Boxin
Zhang, Xiao
Meng, Erli
Feng, Wenjie
Lou, Jian
Ng, See-Kiong
author_facet Li, Lin
Huang, Jiawei
Quan, Qihao
Li, Dan
Li, Boxin
Zhang, Xiao
Meng, Erli
Feng, Wenjie
Lou, Jian
Ng, See-Kiong
contents In this paper, we propose the first VL$\underline{\textbf{M}}$ $\underline{\textbf{a}}$gentic $\underline{\textbf{r}}$easoning framework for few-$\underline{\textbf{s}}$hot multimodal $\underline{\textbf{T}}$ime $\underline{\textbf{S}}$eries $\underline{\textbf{C}}$lassification ($\textbf{MarsTSC}$), which introduces a self-evolving knowledge bank as a dynamic context iteratively refined via reflective agentic reasoning. The framework comprises three collaborative roles: i) Generator conducts reliable classification via reasoning; ii) Reflector diagnoses the root causes of reasoning errors to yield discriminative insights targeting the temporal features overlooked by Generator; iii) Modifier applies verified updates to the knowledge bank to prevent context collapse. We further introduce a test-time update strategy to enable cautious, continuous knowledge bank refinement to mitigate few-shot bias and distribution shift. Extensive experiments across 12 mainstream time series benchmarks demonstrate that $\textbf{MarsTSC}$ delivers substantial and consistent performance gains across 6 VLM backbones, outperforming both classical and foundation model-based time series baselines under few-shot conditions, while producing interpretable rationales that ground each classification decision in human-readable feature evidence.
format Preprint
id arxiv_https___arxiv_org_abs_2605_09395
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Empowering VLMs for Few-Shot Multimodal Time Series Classification via Tailored Agentic Reasoning
Li, Lin
Huang, Jiawei
Quan, Qihao
Li, Dan
Li, Boxin
Zhang, Xiao
Meng, Erli
Feng, Wenjie
Lou, Jian
Ng, See-Kiong
Artificial Intelligence
Machine Learning
Multiagent Systems
Multimedia
I.2.0; I.2.4; I.5.4
In this paper, we propose the first VL$\underline{\textbf{M}}$ $\underline{\textbf{a}}$gentic $\underline{\textbf{r}}$easoning framework for few-$\underline{\textbf{s}}$hot multimodal $\underline{\textbf{T}}$ime $\underline{\textbf{S}}$eries $\underline{\textbf{C}}$lassification ($\textbf{MarsTSC}$), which introduces a self-evolving knowledge bank as a dynamic context iteratively refined via reflective agentic reasoning. The framework comprises three collaborative roles: i) Generator conducts reliable classification via reasoning; ii) Reflector diagnoses the root causes of reasoning errors to yield discriminative insights targeting the temporal features overlooked by Generator; iii) Modifier applies verified updates to the knowledge bank to prevent context collapse. We further introduce a test-time update strategy to enable cautious, continuous knowledge bank refinement to mitigate few-shot bias and distribution shift. Extensive experiments across 12 mainstream time series benchmarks demonstrate that $\textbf{MarsTSC}$ delivers substantial and consistent performance gains across 6 VLM backbones, outperforming both classical and foundation model-based time series baselines under few-shot conditions, while producing interpretable rationales that ground each classification decision in human-readable feature evidence.
title Empowering VLMs for Few-Shot Multimodal Time Series Classification via Tailored Agentic Reasoning
topic Artificial Intelligence
Machine Learning
Multiagent Systems
Multimedia
I.2.0; I.2.4; I.5.4
url https://arxiv.org/abs/2605.09395