mlsanityData to 7 Oct 2026

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

arXiv 2506.13705 ↗First posted 16 Jun 2025cs.LG cs.AIcited by 16

On the map: LLM reasoning → Time series reasoning

also in 1 more
Abstract (arXiv)

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.

Key facts. Cited by 16 papers in the map (data through 7 Oct 2026). Placed in Time series reasoning, one of 17 directions in LLM reasoning. Builds on 44 of its 55 references that are in the map. First posted on arXiv 16 Jun 2025.

Builds on

44 of 55 references.

Closest ancestors · 5

Same direction, newest first

  1. 2025Evaluating System 1 vs. 2 Reasoning Approaches for Zero-Shot Time Series Forecasting: A Benchmark and Insights27 Feb
  2. 2025Achieving Time Series Reasoning Requires Rethinking Model Design, Tasks Formulation, and Evaluation3 Feb
  3. 2024ChatTS: Aligning Time Series with LLMs via Synthetic Data for Enhanced Understanding and Reasoning4 Dec
  4. 2024A Picture is Worth A Thousand Numbers: Enabling LLMs Reason about Time Series via Visualization9 Nov
  5. 2024Language Models Still Struggle to Zero-shot Reason about Time Series17 Apr

Other references · 39

  1. 2017Attention Is All You Needcited by 59,932
  2. 2023GPT-4 Technical Reportcited by 18,132
  3. 2022Training language models to follow instructions with human feedbackcited by 15,712
  4. 2023LLaMA: Open and Efficient Foundation Language Modelscited by 13,958
  5. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,302
  6. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  7. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,393
  8. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 7,916
  9. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,146
  10. 2025Qwen2.5-VL Technical Reportcited by 5,418
  11. 2023Qwen Technical Reportcited by 3,273
  12. 2025DAPO: An Open-Source LLM Reinforcement Learning System at Scalecited by 2,678
27 more
  1. 2020Learning to summarize from human feedbackcited by 2,562
  2. 2020Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecastingcited by 2,141
  3. 2019Fine-Tuning Language Models from Human Preferencescited by 2,124
  4. 2022A Time Series is Worth 64 Words: Long-term Forecasting with Transformerscited by 1,981
  5. 2021Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecastingcited by 1,654
  6. 2022Are Transformers Effective for Time Series Forecasting?cited by 1,400
  7. 2022FEDformer: Frequency Enhanced Decomposed Transformer for Long-term Series Forecastingcited by 1,200
  8. 2025Kimi k1.5: Scaling Reinforcement Learning with LLMscited by 1,193
  9. 2023iTransformer: Inverted Transformers Are Effective for Time Series Forecastingcited by 906
  10. 2023Time-LLM: Time Series Forecasting by Reprogramming Large Language Modelscited by 680
  11. 2023TEST: Text Prototype Aligned Embedding to Activate LLM's Ability for Time Seriescited by 165
  12. 2024CALF: Aligning LLMs for Time Series Forecasting via Cross-modal Fine-Tuningcited by 105
  13. 2024TimeCMA: Towards LLM-Empowered Multivariate Time Series Forecasting via Cross-Modality Alignmentcited by 97
  14. 2024From News to Forecast: Integrating Event Analysis in LLM-Based Time Series Forecasting with Reflectioncited by 93
  15. 2025Time-MQA: Time Series Multi-Task Question Answering with Context Enhancementcited by 74
  16. 2024TSLANet: Rethinking Transformers for Time Series Representation Learningcited by 71
  17. 2021Deep Learning in Human Activity Recognition with Wearable Sensors: A Review on Advancescited by 54
  18. 2025TimeXL: Explainable Multi-modal Time Series Prediction with LLM-in-the-Loopcited by 39
  19. 2022TACTiS: Transformer-Attentional Copulas for Time Seriescited by 38
  20. 2025TimeCAP: Learning to Contextualize, Augment, and Predict Time Series Events with Large Language Model Agentscited by 33
  21. 2024MEIT: Multimodal Electrocardiogram Instruction Tuning on Large Language Models for Report Generationcited by 29
  22. 2024Multi-Modal Forecaster: Jointly Predicting Time Series and Textual Datacited by 20
  23. 2025How Can Time Series Analysis Benefit From Multiple Modalities? A Survey and Outlookcited by 20
  24. 2023Multi-Modal Deep Learning for Credit Rating Prediction Using Text and Numerical Data Streamscited by 13
  25. 2024BjTT: A Large-scale Multimodal Dataset for Traffic Predictioncited by 6
  26. 2024Diverse Intra- and Inter-Domain Activity Style Fusion for Cross-Person Generalization in Activity Recognitioncited by 6
  27. 2020Understanding Electricity-Theft Behavior via Multi-Source Datacited by 5

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 6; 2026: 10.

16 papers.

Sort:

By area:

  1. 2025The Landscape of Agentic Reinforcement Learning for LLMs: A Surveycited by 195
  2. 2025TimeOmni-1: Incentivizing Complex Reasoning with Time Series in Large Language Modelscited by 24
  3. 2025A Survey of Reasoning and Agentic Systems in Time Series with Large Language Modelscited by 21
  4. 2025Reinforcement Learning for Large Model: A Surveycited by 8
  5. 2026STReasoner: Empowering LLMs for Spatio-Temporal Reasoning in Time Series via Spatial-Aware Reinforcement Learningcited by 7
  6. 2025TelecomTS: A Multi-Modal Observability Dataset for Time Series and Language Analysiscited by 5
  7. 2026ARFBench: Benchmarking Time Series Question Answering Ability for Software Incident Responsecited by 5
  8. 2026Time Series Reasoning via Process-Verifiable Thinking Data Synthesis and Scheduling for Tailored LLM Reasoningcited by 3
  9. 2026Adaptive Time Series Reasoning via Segment Selectioncited by 3
  10. 2025Towards Interpretable and Trustworthy Time Series Reasoning: A BlueSky Visioncited by 2
  11. 2026Multi-Modal Time Series Prediction via Mixture of Modulated Expertscited by 1
  12. 2026SenTSR-Bench: Thinking with Injected Knowledge for Time-Series Reasoningcited by 1
4 more
  1. 2026From Consistency to Complementarity: Aligned and Disentangled Multi-modal Learning for Time Series Understanding and Reasoningcited by 0
  2. 2026NeuroCanvas: VLLM-Powered Robust Seizure Detection by Reformulating Multichannel EEG as Imagecited by 0
  3. 2026AnomSeer: Reinforcing Multimodal LLMs to Reason for Time-Series Anomaly Detectioncited by 0
  4. 2026LineupRL: Verifiable Reinforcement Learning for Time Series Captioning via Caption-to-Series Identificationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024Towards Time Series Reasoning with LLMs
  2. 2026TimeThink: Eliciting Compositional Reasoning in Timeseries Large Language Models
  3. 2025Time-R1: Towards Comprehensive Temporal Reasoning in LLMs
  4. 2025TempR1: Improving Temporal Understanding of MLLMs via Temporal-Aware Multi-Task Reinforcement Learning
  5. 2026Thoth: Mid-Training Bridges LLMs to Time Series Understanding
  6. 2025Time Series Forecasting via Reasoning: A Slow-Thinking Approach with Reinforcement Fine-Tuned LLMs
  7. 2025Chat-TS: Enhancing Multi-Modal Reasoning Over Time-Series and Natural Language Data
  8. 2025Can Slow-thinking LLMs Reason Over Time? Empirical Studies in Time Series Forecasting