Elamur Research⌕
☰
home / papers / 2025 / Time Series Forecasting via Reasoning: A Slow-Thinking Approach with Reinforcement Fine-Tuned LLMs

Time Series Forecasting via Reasoning: A Slow-Thinking Approach with Reinforcement Fine-Tuned LLMs

Paper · 2025 · Yitong Zhou, Yucong Luo, Mingyue Cheng et al.

Summary

Most time series forecasting (TSF) models, from ARIMA to Transformers and foundation models, map past values directly to future ones in a single pass. They offer little explicit reasoning, which the authors argue limits how well they adapt to shifting dynamics. Prompting large reasoning LLMs is costly, raises privacy issues and lacks domain alignment. The authors therefore propose Time-R1, which trains an LLM to reason step by step about the series before producing its forecast.

Training has two stages. First, a supervised warm-up teaches the model a reasoning style and output format, using synthetic trajectories generated by DeepSeek-R1. Second, reinforcement learning applies a fine-grained, multi-objective reward designed for forecasting. Its optimizer, GRIP, samples a larger pool of rollouts and keeps them non-uniformly. It also gives extra weight to high-reward forecasts. This counters a weakness of GRPO on graded, non-binary error metrics.

On nine real-world datasets, Time-R1 matches or beats a range of traditional, foundation and LLM-based baselines on most of them, while also producing readable explanations.

Key points

  • Prompt and output format. Prompts combine domain knowledge, timestamps with variable values, and task guidelines. The model writes its reasoning inside <think> tags and the forecast inside <answer> tags.
  • How GRIP works. It draws k·G candidate rollouts and selects G of them by reward-weighted or diversity-preserving sampling. This discourages degenerate outputs such as copying the look-back window or predicting a flat mean line.
  • Evaluation setup. Datasets are ETTh1, ETTh2, ETTm1, ETTm2, Exchange, AQWan, AQShunyi, Wind and NASDAQ. Baselines include PatchTST, iTransformer, TimeMixer, TimesFM-2.5, Chronos-2, GPT4TS, TimeLLM and zero-shot DeepSeek-R1.
  • MAE results (Table 6).
    • Time-R1 leads clearly on ETTh1 (1.2325 vs. 1.4018 for TimeMixer), Exchange (0.0161 vs. 0.0182) and Wind (15.11 vs. 18.10).
    • It trails on AQShunyi and slightly on NASDAQ.
  • Training paradigm comparison.
    • SFT alone tends to copy reasoning superficially.
    • Pure RL yields poorly readable reasoning traces.
    • SFT followed by RL gives the best combination of accuracy and interpretability.
  • Publication and code. The paper appears at CIKM '26, and the code is at github.com/ustc-time-series/Time-R1.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 19,503 worksDecision Sciences · 306 worksSignal Processing · 499 worksManagement Science and Operations Research · 145 works

🧭 Topics 2Research subjects and languages the work is about.Language models & pretraining · 4,145 worksReasoning · 1,948 works

Cite

Cite as Yitong Zhou, Yucong Luo, Mingyue Cheng, Qi Liu, Jiahao Wang, Daoyu Wang and Enhong Chen. 2025. Time Series Forecasting via Reasoning: A Slow-Thinking Approach with Reinforcement Fine-Tuned LLMs. arXiv:2506.10630.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2506.10630 · arXiv 2506.10630

@misc{zhou2025time,
  title = {{Time Series Forecasting via Reasoning: A Slow-Thinking Approach with Reinforcement Fine-Tuned LLMs}},
  author = {Yitong Zhou, Yucong Luo, Mingyue Cheng, Qi Liu, Jiahao Wang, Daoyu Wang and Enhong Chen},
  year = {2025},
  doi = {10.48550/arxiv.2506.10630},
  eprint = {2506.10630},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2506.10630}
}

References

69 references, 19 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Abdul Fatir Ansari, Oleksandr Shchur, Jaris Küken, Andreas Auer, Boran Han, Pedro Mercado, Syama Sundar Rangapuram, Huibin Shen, Lorenzo Stella, Xiyuan Zhang, et al. 2025. Chronos-2: From univariate to universal forecasting. arXiv preprint arXiv:2510.15821 (2025). — [arXiv:2510.15821](https://arxiv.org/abs/2510.15821) find
  2. Vassilis Assimakopoulos and Konstantinos Nikolopoulos. 2000. The theta model: a decomposition approach to forecasting. International journal of forecasting 16, 4 (2000), 521–530. find
  3. George EP Box and Gwilym M Jenkins. 1968. Some recent advances in forecasting and control. Journal of the Royal Statistical Society. Series C (Applied Statistics) 17, 2 (1968), 91–109. find
  4. Ching Chang, Wen-Chih Peng, and Tien-Fu Chen. 2023. Llm4ts: Two-stage fine-tuning for time-series forecasting with pre-trained llms. CoRR (2023). find
  5. Ching Chang, Wei-Yao Wang, Wen-Chih Peng, Tien-Fu Chen, and Sagar Samtani. 2024. Align and Fine-Tune: Enhancing LLMs for Time-Series Forecasting. In NeurIPS Workshop on Time Series in the Age of Large Models. find
  6. Hardy Chen, Haoqin Tu, Fali Wang, Hui Liu, Xianfeng Tang, Xinya Du, Yuyin Zhou, and Cihang Xie. 2025. Sft or rl? an early investigation into training r1-like reasoning large vision-language models. arXiv preprint arXiv:2504.11468 (2025).
  7. Tianqi Chen and Carlos Guestrin. 2016. Xgboost: A scalable tree boosting system. In Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining. 785–794. find
  8. Mingyue Cheng, Xiaoyu Tao, Qi Liu, Hao Zhang, Yiheng Chen, and Defu Lian. 2025. Cross-Domain Pre-training with Language Models for Transferable Time Series Representations. In Proceedings of the Eighteenth ACM International Conference on Web Search and Data Mining. 175–183. find
  9. Mingyue Cheng, Jiahao Wang, Daoyu Wang, Xiaoyu Tao, Qi Liu, and Enhong Chen. 2026. Can slow-thinking llms reason over time? empirical studies in time series forecasting. In Proceedings of the Nineteenth ACM International Conference on Web Search and Data Mining. 99–110. find
  10. Tianzhe Chu, Yuexiang Zhai, Jihan Yang, Shengbang Tong, Saining Xie, Dale Schuurmans, Quoc V Le, Sergey Levine, and Yi Ma. 2025. Sft memorizes, rl generalizes: A comparative study of foundation model post-training. arXiv preprint arXiv:2501.17161 (2025).
  11. Abhimanyu Das, Weihao Kong, Rajat Sen, and Yichen Zhou. 2024. A decoder-only foundation model for time-series forecasting. In Forty-first International Conference on Machine Learning. find
  12. Samuel Dooley, Gurnoor Singh Khurana, Chirag Mohapatra, Siddartha V Naidu, and Colin White. 2023. Forecastpfn: Synthetically-trained zero-shot forecasting. Advances in Neural Information Processing Systems 36 (2023), 2403–2426. find
  13. Fuli Feng, Xiangnan He, Xiang Wang, Cheng Luo, Yiqun Liu, and Tat-Seng Chua. 2019. Temporal relational ranking for stock prediction. ACM Transactions on Information Systems (TOIS) 37, 2 (2019), 1–30. find
  14. Mononito Goswami, Konrad Szafer, Arjun Choudhry, Yifu Cai, Shuo Li, and Artur Dubrawski. 2024. Moment: A family of open time-series foundation models. arXiv preprint arXiv:2402.03885 (2024). — [arXiv:2402.03885](https://arxiv.org/abs/2402.03885) find
  15. Nate Gruver, Marc Finzi, Shikai Qiu, and Andrew G Wilson. 2023. Large language models are zero-shot time series forecasters. Advances in Neural Information Processing Systems 36 (2023), 19622–19635.
  16. Tong Guan, Zijie Meng, Dianqi Li, Shiyu Wang, Chao-Han Huck Yang, Qingsong Wen, Zuozhu Liu, Sabato Marco Siniscalchi, Ming Jin, and Shirui Pan. 2025. Timeomni-1: Incentivizing complex reasoning with time series in large language models. arXiv preprint arXiv:2509.24803 (2025). — [arXiv:2509.24803](https://arxiv.org/abs/2509.24803) find
  17. Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, et al. 2025. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. arXiv preprint arXiv:2501.12948 (2025).
  18. Pradeep Hewage, Ardhendu Behera, Marcello Trovati, Ella Pereira, Morteza Ghahremani, Francesco Palmieri, and Yonghuai Liu. 2020. Temporal convolutional neural (TCN) network for an effective weather forecasting using time-series data from the local weather station. Soft Computing 24 (2020), 16453–164… find
  19. Hansika Hewamalage, Christoph Bergmeir, and Kasun Bandara. 2021. Recurrent neural networks for time series forecasting: Current status and future directions. International Journal of Forecasting 37, 1 (2021), 388–427. find
  20. Rob Hyndman, Anne B Koehler, J Keith Ord, and Ralph D Snyder. 2008. Forecasting with exponential smoothing: the state space approach. Springer Science & Business Media. find
  21. Yushan Jiang, Zijie Pan, Xikun Zhang, Sahil Garg, Anderson Schneider, Yuriy Nevmyvaka, and Dongjin Song. 2024. Empowering time series analysis with large language models: A survey. arXiv preprint arXiv:2402.03182 (2024). — [arXiv:2402.03182](https://arxiv.org/abs/2402.03182)
  22. Ming Jin, Shiyu Wang, Lintao Ma, Zhixuan Chu, James Y Zhang, Xiaoming Shi, Pin-Yu Chen, Yuxuan Liang, Yuan-Fang Li, Shirui Pan, et al. 2023. Time-llm: Time series forecasting by reprogramming large language models. arXiv preprint arXiv:2310.01728 (2023).
  23. Ming Jin, Yifan Zhang, Wei Chen, Kexin Zhang, Yuxuan Liang, Bin Yang, Jindong Wang, Shirui Pan, and Qingsong Wen. 2024. Position: What can large language models tell us about time series analysis. In Forty-first International Conference on Machine Learning.
  24. Leslie Pack Kaelbling, Michael L Littman, and Andrew W Moore. 1996. Reinforcement learning: A survey. Journal of artificial intelligence research 4 (1996), 237–285. find
  25. Timo Kaufmann, Paul Weng, Viktor Bengs, and Eyke Hüllermeier. 2023. A survey of reinforcement learning from human feedback. arXiv preprint arXiv:2312.14925 10 (2023).
  26. Guolin Ke, Qi Meng, Thomas Finley, Taifeng Wang, Wei Chen, Weidong Ma, Qiwei Ye, and Tie-Yan Liu. 2017. Lightgbm: A highly efficient gradient boosting decision tree. Advances in neural information processing systems 30 (2017). find
  27. Yaxuan Kong, Yiyuan Yang, Yoontae Hwang, Wenjie Du, Stefan Zohren, Zhangyang Wang, Ming Jin, and Qingsong Wen. 2025. Time-mqa: Time series multi-task question answering with context enhancement. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Lon…
  28. Leander Kotzur, Peter Markewitz, Martin Robinius, and Detlef Stolten. 2018. Time series aggregation for energy system design: Modeling seasonal storage. Applied Energy 213 (2018), 123–135. find
  29. Guokun Lai, Wei-Cheng Chang, Yiming Yang, and Hanxiao Liu. 2018. Modeling long-and short-term temporal patterns with deep neural networks. In The 41st international ACM SIGIR conference on research & development in information retrieval. 95–104. find
  30. Jiaxiang Li, Siliang Zeng, Hoi-To Wai, Chenliang Li, Alfredo Garcia, and Mingyi Hong. 2024. Getting more juice out of the sft data: Reward learning from human demonstration improves sft for llm alignment. Advances in Neural Information Processing Systems 37 (2024), 124292–124318. find
  31. Li Li, Xiaonan Su, Yi Zhang, Yuetong Lin, and Zhiheng Li. 2015. Trend modeling for traffic time series analysis: An integrated study. IEEE Transactions on Intelligent Transportation Systems 16, 6 (2015), 3430–3439. find
  32. Shiyang Li, Xiaoyong Jin, Yao Xuan, Xiyou Zhou, Wenhu Chen, Yu-Xiang Wang, and Xifeng Yan. 2019. Enhancing the locality and breaking the memory bottleneck of transformer on time series forecasting. Advances in neural information processing systems 32 (2019). find
  33. Haoxin Liu, Zhiyuan Zhao, Shiduo Li, and B Aditya Prakash. 2025. Evaluating system 1 vs. 2 reasoning approaches for zero-shot time series forecasting: A benchmark and insights. arXiv preprint arXiv:2503.01895 (2025). find
  34. Haoxin Liu, Zhiyuan Zhao, Jindong Wang, Harshavardhan Kamarthi, and B Aditya Prakash. 2024. Lstprompt: Large language models as zero-shot time series forecasters by long-short-term prompting. arXiv preprint arXiv:2402.16132 (2024). find
  35. Peiyuan Liu, Hang Guo, Tao Dai, Naiqi Li, Jigang Bao, Xudong Ren, Yong Jiang, and Shu-Tao Xia. 2025. Calf: Aligning llms for time series forecasting via cross-modal fine-tuning. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 39. 18915–18923. find
  36. Qingxiang Liu, Xu Liu, Chenghao Liu, Qingsong Wen, and Yuxuan Liang. 2024. Time-ffm: Towards lm-empowered federated foundation model for time series forecasting. arXiv preprint arXiv:2405.14252 (2024). — [arXiv:2405.14252](https://arxiv.org/abs/2405.14252) find
  37. Yong Liu, Tengge Hu, Haoran Zhang, Haixu Wu, Shiyu Wang, Lintao Ma, and Mingsheng Long. 2023. itransformer: Inverted transformers are effective for time series forecasting. arXiv preprint arXiv:2310.06625 (2023). — [arXiv:2310.06625](https://arxiv.org/abs/2310.06625) find
  38. Junkai Lu, Peng Chen, Xingjian Wu, Yang Shu, Chenjuan Guo, Christian S Jensen, and Bin Yang. 2026. PATRA: Pattern-Aware Alignment and Balanced Reasoning for Time Series Question Answering. arXiv preprint arXiv:2602.23161 (2026). — [arXiv:2602.23161](https://arxiv.org/abs/2602.23161) find
  39. Yu Meng, Mengzhou Xia, and Danqi Chen. 2024. Simpo: Simple preference optimization with a reference-free reward. Advances in Neural Information Processing Systems 37 (2024), 124198–124235. find
  40. Md Mahmuddun Nabi Murad, Mehmet Aktukmak, and Yasin Yilmaz. 2025. Wpmixer: Efficient multi-resolution mixing for long-term time series forecasting. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 39. 19581–19588. find
  41. Juntong Ni, Shiyu Wang, Ming Jin, Qi He, and Wei Jin. 2026. STReasoner: Empowering LLMs for Spatio-Temporal Reasoning in Time Series via Spatial-Aware Reinforcement Learning. arXiv preprint arXiv:2601.03248 (2026).
  42. Yuqi Nie, Nam H Nguyen, Phanwadee Sinthong, and Jayant Kalagnanam. 2022. A time series is worth 64 words: Long-term forecasting with transformers. arXiv preprint arXiv:2211.14730 (2022).
  43. Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. 2022. Training language models to follow instructions with human feedback. Advances in neural information processing systems 35 (2022), 27730–27744.
  44. Richard Y Pang, Weizhe Yuan, Kyunghyun Cho, He He, Sainbayar Sukhbaatar, and Jason Weston. 2024. Iterative reasoning preference optimization. Advances in Neural Information Processing Systems 37 (2024), 116617–116637.
  45. Rafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D Manning, Stefano Ermon, and Chelsea Finn. 2023. Direct preference optimization: Your language model is secretly a reward model. Advances in Neural Information Processing Systems 36 (2023), 53728–53741.
  46. Kady Sako, Berthine Nyunga Mpinda, and Paulo Canas Rodrigues. 2022. Neural networks for financial time series forecasting. Entropy 24, 5 (2022), 657. find
  47. David Salinas, Valentin Flunkert, Jan Gasthaus, and Tim Januschowski. 2020. DeepAR: Probabilistic forecasting with autoregressive recurrent networks. International journal of forecasting 36, 3 (2020), 1181–1191. find
  48. Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, YK Li, Y Wu, et al. 2024. Deepseekmath: Pushing the limits of mathematical reasoning in open language models. arXiv preprint arXiv:2402.03300 (2024).
  49. Guangming Sheng, Chi Zhang, Zilingfeng Ye, Xibin Wu, Wang Zhang, Ru Zhang, Yanghua Peng, Haibin Lin, and Chuan Wu. 2024. Hybridflow: A flexible and efficient rlhf framework. arXiv preprint arXiv:2409.19256 (2024).
  50. Huatong Song, Jinhao Jiang, Yingqian Min, Jie Chen, Zhipeng Chen, Wayne Xin Zhao, Lei Fang, and Ji-Rong Wen. 2025. R1-searcher: Incentivizing the search capability in llms via reinforcement learning. arXiv preprint arXiv:2503.05592 (2025).
  51. Mingtian Tan, Mike Merrill, Vinayak Gupta, Tim Althoff, and Tom Hartvigsen. 2024. Are language models actually useful for time series forecasting? Advances in Neural Information Processing Systems 37 (2024), 60162–60191. find
  52. Xiaoyu Tao, Mingyue Cheng, Ze Guo, Shuo Yu, Yaguo Liu, Qi Liu, and Shijin Wang. 2026. Memcast: Memory-driven time series forecasting with experience-conditioned reasoning. arXiv preprint arXiv:2602.03164 (2026). — [arXiv:2602.03164](https://arxiv.org/abs/2602.03164) find
  53. Xiaoyu Tao, Mingyue Cheng, Chuang Jiang, Tian Gao, Huanjian Zhang, and Yaguo Liu. 2026. Cast-R1: Learning tool-augmented sequential decision policies for time series forecasting. arXiv preprint arXiv:2602.13802 (2026). — [arXiv:2602.13802](https://arxiv.org/abs/2602.13802) find
  54. Jiahao Wang, Mingyue Cheng, Qingyang Mao, Yitong Zhou, Daoyu Wang, Qi Liu, Feiyang Xu, and Xin Li. 2025. Tabletime: Reformulating time series classification as training-free table understanding with large language models. In Proceedings of the 34th ACM International Conference on Information and Kno… find
  55. Shiyu Wang, Haixu Wu, Xiaoming Shi, Tengge Hu, Huakun Luo, Lintao Ma, James Y Zhang, and Jun Zhou. 2024. Timemixer: Decomposable multiscale mixing for time series forecasting. arXiv preprint arXiv:2405.14616 (2024). — [arXiv:2405.14616](https://arxiv.org/abs/2405.14616) find
  56. Yilin Wang, Peixuan Lei, Jie Song, Yuzhe Hao, Tao Chen, Yuxuan Zhang, Lei Jia, Yuanxiang Li, and Zhongyu Wei. 2025. Itformer: Bridging time series and natural language for multi-modal qa with large-scale multitask dataset. arXiv preprint arXiv:2506.20093 (2025). — [arXiv:2506.20093](https://arxiv.org/abs/2506.20093) find
  57. Yuxuan Wang, Haixu Wu, Jiaxiang Dong, Guo Qin, Haoran Zhang, Yong Liu, Yunzhong Qiu, Jianmin Wang, and Mingsheng Long. 2024. Timexer: Empowering transformers for time series forecasting with exogenous variables. arXiv preprint arXiv:2402.19072 (2024). find
  58. Qingsong Wen, Zhe Zhang, Yan Li, and Liang Sun. 2020. Fast RobustSTL: Efficient and robust seasonal-trend decomposition for time series with complex patterns. In Proceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining. 2203–2213. find
  59. Qingsong Wen, Tian Zhou, Chaoli Zhang, Weiqi Chen, Ziqing Ma, Junchi Yan, and Liang Sun. 2023. Transformers in Time Series: A Survey. Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence, IJCAI-23, 6778–6786. find
  60. Haixu Wu, Jiehui Xu, Jianmin Wang, and Mingsheng Long. 2021. Autoformer: Decomposition transformers with auto-correlation for long-term series forecasting. Advances in neural information processing systems 34 (2021), 22419–22430. find
  61. Qiying Yu, Zheng Zhang, Ruofei Zhu, Yufeng Yuan, Xiaochen Zuo, Yu Yue, Weinan Dai, Tiantian Fan, Gaohong Liu, Lingjun Liu, et al. 2026. Dapo: An open-source llm reinforcement learning system at scale. Advances in Neural Information Processing Systems 38 (2026), 113222–113244.
  62. Ailing Zeng, Muxi Chen, Lei Zhang, and Qiang Xu. 2023. Are transformers effective for time series forecasting?. In Proceedings of the AAAI conference on artificial intelligence, Vol. 37. 11121–11128. find
  63. G Peter Zhang. 2003. Time series forecasting using a hybrid ARIMA and neural network model. Neurocomputing 50 (2003), 159–175. find
  64. Shuyi Zhang, Bin Guo, Anlan Dong, Jing He, Ziping Xu, and Song Xi Chen. 2017. Cautionary tales on air-quality improvement in Beijing. Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 473, 2205 (2017), 20170457. find
  65. Xiyuan Zhang, Ranak Roy Chowdhury, Rajesh K Gupta, and Jingbo Shang. 2024. Large language models for time series: A survey. arXiv preprint arXiv:2402.01801 (2024). — [arXiv:2402.01801](https://arxiv.org/abs/2402.01801) find
  66. Yunhao Zhang and Junchi Yan. 2023. Crossformer: Transformer utilizing cross-dimension dependency for multivariate time series forecasting. In The eleventh international conference on learning representations. find
  67. Chujie Zheng, Shixuan Liu, Mingze Li, Xiong-Hui Chen, Bowen Yu, Chang Gao, Kai Dang, Yuqiong Liu, Rui Men, An Yang, et al. 2025. Group sequence policy optimization. arXiv preprint arXiv:2507.18071 (2025).
  68. Haoyi Zhou, Shanghang Zhang, Jieqi Peng, Shuai Zhang, Jianxin Li, Hui Xiong, and Wancai Zhang. 2021. Informer: Beyond efficient transformer for long sequence time-series forecasting. In Proceedings of the AAAI conference on artificial intelligence, Vol. 35. 11106–11115. find
  69. Tian Zhou, Peisong Niu, Liang Sun, Rong Jin, et al. 2023. One fits all: Power general time series analysis by pretrained lm. Advances in neural information processing systems 36 (2023), 43322–43355. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zhou-2025-time-series-forecasting-reasoning-slow/