Elamur Research⌕
☰
home / papers / 2026 / Inferring Events from Time Series using Language Models

Inferring Events from Time Series using Language Models

Paper · 2026 · Mingtian Tan, Mike A Merrill, Zachary Gottesman et al.

Summary

The paper asks whether large language models can work backwards from a numeric time series to the real-world events that likely produced it. Earlier benchmarks either covered purely numeric tasks such as forecasting and anomaly detection, or event reasoning with no numeric signal. The authors fill this gap with GAMETime, a benchmark generated programmatically from sports data. It pairs ESPN win-probability curves with play-by-play descriptions from about 4,200 NBA and NFL games. In each task, part of the event log is hidden, and the model picks which candidate event sequence best fits the observed probability changes. Because sports data keeps updating, the benchmark can be refreshed regularly, which reduces contamination risk. Across 18 LLMs, reasoning models did best, and OpenAI's o1 led. The authors also post-trained a small Qwen2.5 1.5B model with distillation followed by GRPO-based reinforcement learning, which brought it close to o1. The approach transfers to finance, trade, health and energy series, so it offers a general way to test whether models link numeric data to the events behind it.

Key points

  • Dataset size: about 1.7 million timesteps of values plus events from NBA and NFL games. Code and data are released as GAMETime on GitHub.
  • NBA accuracy:
    • o1: 83%
    • DS-R1-distill-Qwen-32B: 68%
    • GPT4o: 41%
  • Post-training result: after distillation and GRPO, Qwen2.5 1.5B rose from the weakest model to second place, behind only o1.
  • Ablation findings:
    • Negative options built by reordering the true events were easier to reject, suggesting models use sequential logic.
    • Accuracy rose slightly when the distractor's time series was less similar to the target.
  • Other domains: the method was applied to Time-MMD (IMEX trade, influenza, gasoline) and CryptoTrade Bitcoin data.
    • Masking numbers inside the event texts lowered scores, but GPT-4o still exceeded 50% in every domain.
    • A noted limitation is that ESPN win probabilities are model estimates rather than ground truth.

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

 Institutions

🎓Stanford UniversityStanford · United States434 works · 314,419 citations🎓University of WashingtonSeattle · United States353 works · 200,043 citations🎓University of VirginiaCharlottesville · United States36 works · 9,730 citations

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksSignal Processing · 452 works

🧭 Topics 1Research subjects and languages the work is about.Language models & pretraining · 3,320 works

Cite

Cite as Mingtian Tan, Mike A Merrill, Zachary Gottesman, Tim Althoff, David Evans and Thomas Hartvigsen. 2026. Inferring Events from Time Series using Language Models. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-long.157

@inproceedings{tan2026inferring,
  title = {{Inferring Events from Time Series using Language Models}},
  author = {Mingtian Tan, Mike A Merrill, Zachary Gottesman, Tim Althoff, David Evans and Thomas Hartvigsen},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.157},
  url = {https://doi.org/10.18653/v1/2026.acl-long.157}
}

References

53 references, 18 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Marah Abdin, Jyoti Aneja, Harkirat Behl, et al. 2024. Phi-4 technical report. arXiv:2412.08905.
  2. Josh Achiam, Steven Adler, Sandhini Agarwal, et al. 2023. GPT-4 technical report. arXiv:2303.08774.
  3. Yejin Bang, Samuel Cahyawijaya, Nayeon Lee, Wenliang Dai, Dan Su, Bryan Wilie, Holy Lovenia, Ziwei Ji, Tiezheng Yu, Willy Chung, Quyet V. Do, Yan Xu, and Pascale Fung. 2023. A multitask, multilingual, multimodal evaluation of ChatGPT on reasoning, hallucination, and interactivity. In 13 th Internati…
  4. Defu Cao, Furong Jia, Sercan O Arik, Tomas Pfister, Yixiang Zheng, Wen Ye, and Yan Liu. 2024. Tempo: Prompt-based generative pre-trained transformer for time series forecasting. In International Conference on Learning Representations (ICLR). find
  5. Nimeesha Chan, Felix Parker, William Bennett, Tianyi Wu, Mung Yao Jia, James Fackler, and Kimia Ghobadi. 2024. MedTsLLM: Leveraging LLMs for multimodal medical time series analysis. In 9 th Machine Learning for Healthcare Conference. find
  6. Junyan Cheng and Peter Chin. 2024. Sociodojo: Building lifelong analytical agents with real-world text and time series. In International Conference on Learning Representations (ICLR). find
  7. Zheng Chu, Jingchang Chen, Qianglong Chen, Weijiang Yu, Haotian Wang, Ming Liu, and Bing Qin. 2024. TimeBench: A comprehensive evaluation of temporal reasoning abilities in large language models. In 62 nd Annual Meeting of the Association for Computational Linguistics (ACL). find
  8. Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, Christopher Hesse, and John Schulman. 2021. Training Verifiers to Solve Math Word Problems. arXiv:2110.14168.
  9. DeepSeek-AI. 2025. DeepSeek-R1: Incentivizing reasoning capability in LLMs via reinforcement learning. arXiv:2501.12948.
  10. Manqing Dong, Hao Huang, and Longbing Cao. 2025. Can LLMs serve as time series anomaly detectors? In International Conference on Learning Representations (ICLR). find
  11. Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, et al. 2024. The Llama 3 herd of models. arXiv:2407.21783.
  12. Bahare Fatemi, Mehran Kazemi, Anton Tsitsulin, Karishma Malkan, Jinyeong Yim, John Palowitch, Sungyong Seo, Jonathan Halcrow, and Bryan Perozzi. 2025. Test of Time: A benchmark for evaluating LLMs on temporal reasoning. In International Conference on Learning Representations (ICLR). find
  13. Daniel Garcia-Gonzalez, Daniel Rivero, Enrique Fernandez-Blanco, and Miguel R. Luaces. 2020. A public domain dataset for human activity recognition using smartphones. Sensors, 20(8). find
  14. BIG bench authors. 2023. Beyond the imitation game: Quantifying and extrapolating the capabilities of language models. Transactions on Machine Learning Research.
  15. Hrayr Harutyunyan, Hrant Khachatrian, David C Kale, Greg Ver Steeg, and Aram Galstyan. 2019. Multitask learning and benchmarking with clinical time series data. Scientific data, 6(1):96. find
  16. Yifu Cai, Arjun Choudhry, Mononito Goswami, and Artur Dubrawski. 2024. TimeSeriesExam: A time series understanding exam. In NeurIPS Workshop on Time Series in the Age of Large Models. find
  17. Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. 2020. Measuring massive multitask language understanding. arXiv:2009.03300.
  18. Hugging Face. 2025. Open R1: A fully open reproduction of DeepSeek-R1. https://github.com/ huggingface/open-r1. ↗
  19. Pranab Islam, Anand Kannappan, Douwe Kiela, Rebecca Qian, Nino Scherrer, and Bertie Vidgen. 2023. Financebench: A new benchmark for financial question answering. arXiv:2311.11944. — [arXiv:2311.11944](https://arxiv.org/abs/2311.11944) find
  20. Naman Jain, King Han, Alex Gu, Wen-Ding Li, Fanjia Yan, Tianjun Zhang, Sida Wang, Armando Solar-Lezama, Koushik Sen, and Ion Stoica. 2024. Livecodebench: Holistic and contamination free evaluation of large language models for code. arXiv:2403.07974.
  21. Albert Q Jiang, Alexandre Sablayrolles, Antoine Roux, et al. 2024a. Mixtral of Experts. arXiv:2401.04088. find
  22. Yushan Jiang, Zijie Pan, Xikun Zhang, Sahil Garg, Anderson Schneider, Yuriy Nevmyvaka, and Dongjin Song. 2024b. Empowering time series analysis with large language models: A survey. arXiv:2402.03182. — [arXiv:2402.03182](https://arxiv.org/abs/2402.03182)
  23. Ming Jin, Yifan Zhang, Wei Chen, Kexin Zhang, Yuxuan Liang, Bin Yang, Jindong Wang, Shirui Pan, and Qingsong Wen. 2024. Position paper: What can large language models tell us about time series analysis. In International Conference on Machine Learning (ICML). find
  24. Ezra Karger, Houtan Bastani, Chen Yueh-Han, Zachary Jacobs, Danny Halawi, Fred Zhang, and Philip E. Tetlock. 2025. ForecastBench: A dynamic benchmark of AI forecasting capabilities. In International Conference on Learning Representations (ICLR). find
  25. Svenja Kenneweg, Jörg Deigmöller, Philipp Cimiano, and Julian Eggert. 2025. Traveler: A benchmark for evaluating temporal reasoning across vague, implicit and explicit references. arXiv:2505.01325. — [arXiv:2505.01325](https://arxiv.org/abs/2505.01325) find
  26. Hoyoung Lee, Youngsoo Choi, and Yuhee Kwon. 2024. Quantifying qualitative insights: Leveraging LLMs to market predict. arXiv:2411.08404. — [arXiv:2411.08404](https://arxiv.org/abs/2411.08404) find
  27. Yuan Li, Bingqiao Luo, Qian Wang, Nuo Chen, Xu Liu, and Bingsheng He. 2024a. CryptoTrade: A reflective LLM-based agent to guide zero-shot cryptocurrency trading. In Conference on Empirical Methods in Natural Language Processing (EMNLP). find
  28. Jun Liu, Chaoyun Zhang, Jiaxu Qian, Minghua Ma, Si Qin, Chetan Bansal, Qingwei Lin, Saravan Rajmohan, and Dongmei Zhang. 2024b. Large language models can deliver accurate and interpretable time series anomaly detection. arXiv:2405.15370. — [arXiv:2405.15370](https://arxiv.org/abs/2405.15370) find
  29. Zichen Liu, Changyu Chen, Wenjun Li, Penghui Qi, Tianyu Pang, Chao Du, Wee Sun Lee, and Min Lin. 2025. Understanding r1-zero-like training: A critical perspective. In Conference on Language Modeling (COLM).
  30. Sara Vera Marjanović, Arkil Patel, Vaibhav Adlakha, et al. 2025. DeepSeek-R1 Thoughtology: Let’s think about LLM reasoning. arXiv:2504.07128. find
  31. Mike A Merrill, Mingtian Tan, Vinayak Gupta, Thomas Hartvigsen, and Tim Althoff. 2024. Language models still struggle to zero-shot reason about time series. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing (EMNLP). find
  32. Niklas Muennighoff, Zitong Yang, Weijia Shi, Xiang Lisa Li, Li Fei-Fei, Hannaneh Hajishirzi, Luke Zettlemoyer, Percy Liang, Emmanuel Candès, and Tatsunori Hashimoto. 2025. s1: Simple test-time scaling. In Conference on Empirical Methods in Natural Language Processing (EMNLP).
  33. OpenAI. 2024. Learning to reason with LLMs. find
  34. Ankit Pal, Logesh Kumar Umapathi, and Malaikannan Sankarasubbu. 2022. MedMCQA: A large-scale multi-subject multi-choice dataset for medical domain question answering. In Conference on Health, Inference, and Learning.
  35. Yinzhu Quan and Zefang Liu. 2024. EconLogicQA: A question-answering benchmark for evaluating large language models in economic sequential reasoning. In Conference on Empirical Methods in Natural Language Processing (EMNLP). find
  36. Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Mingchuan Zhang, Y. Wu Y.K. Li, and Daya Guo. DeepSeekMath: Pushing the limits of mathematical reasoning in open language models. arXiv:2402.03300.
  37. Zechen Li, Shohreh Deldari, Linyao Chen, Hao Xue, and Flora D Salim. 2024b. Sensorllm: Aligning large language models with motion sensors for human activity recognition. arXiv:2410.10624.
  38. Haozhan Shen, Peng Liu, Jingcheng Li, Chunxin Fang, Yibo Ma, Jiajia Liao, Qiaoli Shen, Zilun Zhang, Kangjia Zhao, Qianqian Zhang, Ruochen Xu, and Tiancheng Zhao. 2025. VLM-R1: A stable and generalizable R1-style large vision-language model. arXiv:2504.07615. find
  39. Haoxin Liu, Shangqing Xu, Zhiyuan Zhao, Lingkai Kong, Harshavardhan Kamarthi, Aditya B Sasanur, Megha Sharma, Jiaming Cui, Qingsong Wen, Chao Zhang, et al. 2024a. Time-MMD: Multi-domain multimodal dataset for time series analysis. In 38 th Conference on Neural Information Processing Systems Datasets… find
  40. Mirac Suzgun, Nathan Scales, Nathanael Schärli, Sebastian Gehrmann, Yi Tay, Hyung Won Chung, Aakanksha Chowdhery, Quoc Le, Ed Chi, Denny Zhou, and Jason Wei. 2023. Challenging BIG-bench tasks and whether chain-of-thought can solve them. In Findings of the Association for Computational Linguistics: A…
  41. Chengsen Wang, Qi Qi, Jingyu Wang, Haifeng Sun, Zirui Zhuang, Jinming Wu, Lei Zhang, and Jianxin Liao. 2025. ChatTime: A unified multimodal time series foundation model bridging numerical and textual data. In AAAI Conference on Artificial Intelligence. find
  42. Shengkun Wang, Taoran Ji, Linhan Wang, Yanshen Sun, Shang-Ching Liu, Amit Kumar, and Chang-Tien Lu. 2024a. StockTime: A time series specialized large language model architecture for stock price prediction. arXiv:2409.08281. — [arXiv:2409.08281](https://arxiv.org/abs/2409.08281) find
  43. Xinlei Wang, Maike Feng, Jing Qiu, Jinjin Gu, and Junhua Zhao. 2024b. From news to forecast: Integrating event analysis in llm-based time series forecasting with reflection. In Advances in Neural Information Processing Systems (NeurIPS). find
  44. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in Neural Information Processing Systems (NeurIPS).
  45. Colin White, Samuel Dooley, Manley Roberts, et al. 2024. LiveBench: A challenging, contamination-free LLM benchmark. arXiv:2406.19314. find
  46. Andrew Robert Williams, Arjun Ashok, Étienne Marcotte, Valentina Zantedeschi, Jithendaraa Subramanian, Roland Riachi, James Requeima, Alexandre Lacoste, Irina Rish, Nicolas Chapados, and Alexandre Drouin. 2025. Context is key: A benchmark for forecasting with essential textual information. In 42 nd… find
  47. Qianqian Xie, Weiguang Han, Xiao Zhang, Yanzhao Lai, Min Peng, Alejandro Lopez-Lira, and Jimin Huang. 2023. PIXIU: A large language model, instruction data and evaluation benchmark for finance. arXiv:2306.05443. — [arXiv:2306.05443](https://arxiv.org/abs/2306.05443) find
  48. Zhe Xie, Zeyan Li, Xiao He, Longlong Xu, Xidao Wen, Tieying Zhang, Jianjun Chen, Rui Shi, and Dan Pei. 2025. ChatTS: Aligning time series with llms via synthetic data for enhanced understanding and reasoning. In Proceedings of the VLDB Endowment, 2025. find
  49. Siheng Xiong, Ali Payani, Ramana Kompella, and Faramarz Fekri. 2024. Large language models can learn temporal reasoning. In 62 nd Annual Meeting of the Association for Computational Linguistics (ACL). find
  50. Zhijian Xu, Yuxuan Bian, Jianyuan Zhong, Xiangyu Wen, and Qiang Xu. 2024. Beyond trend and periodicity: Guiding time series forecasting with textual cues. arXiv:2405.13522. — [arXiv:2405.13522](https://arxiv.org/abs/2405.13522) find
  51. Rosie Zhao, Alexandru Meterez, Sham Kakade, Cengiz Pehlevan, Samy Jelassi, and Eran Malach. 2025. Echo chamber: Rl post-training amplifies behaviors learned in pretraining. arXiv:2504.07912. find
  52. Zihao Zhou and Rose Yu. 2025. Can LLMs understand time series anomalies? In International Conference on Learning Representations (ICLR). find
  53. Terry Yue Zhuo, Minh Chien Vu, Jenny Chim, Han Hu, Wenhao Yu, Ratnadira Widyasari, Imam Nur Bani Yusuf, Haolan Zhan, Junda He, Indraneil Paul, et al. 2025. BigCodeBench: Benchmarking code generation with diverse function calls and complex instructions. In International Conference on Learning Represe…

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/tan-2026-inferring-events-time-series-language/