TimeMaster: Training Time-Series Multimodal LLMs to Reason via Reinforcement Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Junru, Feng, Lang, Guo, Xu, Wu, Yuhan, Dong, Yabo, Xu, Duanqing
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908409103646720
author Zhang, Junru
Feng, Lang
Guo, Xu
Wu, Yuhan
Dong, Yabo
Xu, Duanqing
author_facet Zhang, Junru
Feng, Lang
Guo, Xu
Wu, Yuhan
Dong, Yabo
Xu, Duanqing
contents Time-series reasoning remains a significant challenge in multimodal large language models (MLLMs) due to the dynamic temporal patterns, ambiguous semantics, and lack of temporal priors. In this work, we introduce TimeMaster, a reinforcement learning (RL)-based method that enables time-series MLLMs to perform structured, interpretable reasoning directly over visualized time-series inputs and task prompts. TimeMaster adopts a three-part structured output format, reasoning, classification, and domain-specific extension, and is optimized via a composite reward function that aligns format adherence, prediction accuracy, and open-ended insight quality. The model is trained using a two-stage pipeline: we first apply supervised fine-tuning (SFT) to establish a good initialization, followed by Group Relative Policy Optimization (GRPO) at the token level to enable stable and targeted reward-driven improvement in time-series reasoning. We evaluate TimeMaster on the TimerBed benchmark across six real-world classification tasks based on Qwen2.5-VL-3B-Instruct. TimeMaster achieves state-of-the-art performance, outperforming both classical time-series models and few-shot GPT-4o by over 14.6% and 7.3% performance gain, respectively. Notably, TimeMaster goes beyond time-series classification: it also exhibits expert-like reasoning behavior, generates context-aware explanations, and delivers domain-aligned insights. Our results highlight that reward-driven RL can be a scalable and promising path toward integrating temporal understanding into time-series MLLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2506_13705
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle TimeMaster: Training Time-Series Multimodal LLMs to Reason via Reinforcement Learning
Zhang, Junru
Feng, Lang
Guo, Xu
Wu, Yuhan
Dong, Yabo
Xu, Duanqing
Machine Learning
Artificial Intelligence
Time-series reasoning remains a significant challenge in multimodal large language models (MLLMs) due to the dynamic temporal patterns, ambiguous semantics, and lack of temporal priors. In this work, we introduce TimeMaster, a reinforcement learning (RL)-based method that enables time-series MLLMs to perform structured, interpretable reasoning directly over visualized time-series inputs and task prompts. TimeMaster adopts a three-part structured output format, reasoning, classification, and domain-specific extension, and is optimized via a composite reward function that aligns format adherence, prediction accuracy, and open-ended insight quality. The model is trained using a two-stage pipeline: we first apply supervised fine-tuning (SFT) to establish a good initialization, followed by Group Relative Policy Optimization (GRPO) at the token level to enable stable and targeted reward-driven improvement in time-series reasoning. We evaluate TimeMaster on the TimerBed benchmark across six real-world classification tasks based on Qwen2.5-VL-3B-Instruct. TimeMaster achieves state-of-the-art performance, outperforming both classical time-series models and few-shot GPT-4o by over 14.6% and 7.3% performance gain, respectively. Notably, TimeMaster goes beyond time-series classification: it also exhibits expert-like reasoning behavior, generates context-aware explanations, and delivers domain-aligned insights. Our results highlight that reward-driven RL can be a scalable and promising path toward integrating temporal understanding into time-series MLLMs.
title TimeMaster: Training Time-Series Multimodal LLMs to Reason via Reinforcement Learning
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2506.13705