Pith. sign in

REVIEW 4 major objections 5 minor 154 references

CastFSR: A Fast--Slow--Reflect Agentic Reasoning Framework for Context-Aware Time Series Forecasting

T0 review · 4 major / 5 minor · reviewed 2026-08-08 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read CastFSR claims that context-aware forecasting is best reformulated as a Fast-Slow-Reflect agentic workflow, in which numerical priors, contextual reasoning, and reflective consistency checks are orchestrated as sequential decisions.

desk verdict Useful agentic forecasting framework whose headline gains are not yet attributable to the workflow itself, because Chronos-2 sits inside the system and never appears as a baseline. read the letter →

arxiv 2608.03031 v1 pith:K3A6PRQ5 submitted 2026-08-04 cs.AI

Xiaoyu Tao , Mingyue Cheng , Bokai Pan , Chuang Jiang , Huanjian Zhang , Tian Gao , Yaguo Liu , Qi Liu
show 1 more author
Enhong Chen
This is my paper · ORCID
classification cs.AI
keywords timeseriesforecastingcontext-awareagenticreasoninglargelanguagemodelsforecastpriorreinforcementlearningFast-Slow-Reflectenergy
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

CastFSR claims that context-aware time series forecasting is best treated as an agentic decision process, not a single mapping from history to future values. The paper proposes a three-stage workflow: fast thinking builds a data-driven forecast prior by routing the input to a lightweight statistical, deep, or foundation forecaster; slow deliberation retrieves relevant contextual evidence and reasons about how it reshapes future dynamics; reflection checks the candidate forecast against temporal, contextual, and domain constraints and corrects localized violations. This orchestration is meant to keep the numerical stability of mature forecasters while letting a language model decide when context should override them. If correct, the workflow improves accuracy and constraint consistency across energy benchmarks while remaining instantiable either with off-the-shelf LLMs or a compact model fine-tuned in two stages.

What carries the argument

The central object is the Fast-Slow-Reflect policy, a three-stage sequential decision loop. In fast thinking, diagnostic tools profile the series and a forecaster pool supplies a numerical prior; in slow deliberation, a context library with adaptive look-back windows supplies evidence used to revise the prior; in reflective evaluation, consistency checks decide whether to accept, locally correct, or re-enter earlier stages. The same workflow is model-agnostic: an off-the-shelf LLM can execute it directly, and supervised fine-tuning followed by multi-turn reinforcement learning transfers it to a compact model.

What would settle it

Re-run the same benchmarks giving every baseline the identical future-known covariates (weather forecasts, load forecasts, and exogenous series) with the same look-back and horizon settings; if CastFSR's advantage shrinks or reverses, the reported gains come from information access rather than from the Fast-Slow-Reflect workflow.

Watch

Extended reading notes

Core claim

The paper's central claim is that the Fast-Slow-Reflect workflow yields forecasts that are both more accurate and more constraint-consistent than existing LLM-based, deep-learning, and statistical baselines. The key move is to split responsibilities: the language model is not a numerical predictor but an orchestrator that selects a forecaster, interprets context, and audits the result. CastFSR-Zero, the training-free version, achieves best or second-best performance on most metrics across the ETT, Wind, and EPF benchmarks, and CastFSR-R1, the compact post-trained version, further improves those results on most benchmarks. The paper argues that these results validate the sequential decision-making formulation of context-aware forecasting.

Load-bearing premise

The central comparison depends on all baselines having access to the same future-known contextual covariates under the same protocol, and the paper does not specify that this was the case.

Editorial extensions

If this is right

  • Forecasting quality improves most when the forecast prior comes from a specialized numerical model rather than from LLM-generated numbers; removing the fast-thinking stage causes the largest degradation in the paper's ablation.
  • Contextual evidence should be used selectively: the agent preserves the prior when evidence is weak or redundant and applies targeted revisions only when context aligns with the forecast horizon.
  • Reflective checks catch physically implausible outputs, such as negative wind-power forecasts, and repair them without regenerating the entire horizon.
  • The two-stage SFT-plus-RL training lets a compact 4B-class model internalize the orchestration, with the full CastFSR-R1 reaching the best or tied-best performance on most benchmarks.
  • The workflow generalizes across coordinator LLMs: no single pretrained model dominates, and the framework's performance is model-agnostic.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper does not report an information-access audit; a controlled rerun in which every baseline receives exactly the same future-known covariates would separate the value of the workflow from the value of extra inputs.
  • A natural extension is to swap the forecaster pool for a different family of numerical models: if accuracy moves only with prior quality and not with the reasoning stage, the orchestration's role is confirmed as routing rather than generation.
  • The reflection stage's domain constraints could be tested on tasks with hard operational limits, such as grid capacity or inventory bounds, where consistency violations are cheap to measure.
  • Because the compact post-trained model outperforms the training-free version on most benchmarks, further RL iterations on harder multi-step trajectories could probe how closely a small model can approach a large coordinator.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes CastFSR, an agentic framework for context-aware time series forecasting organized as a Fast–Slow–Reflect workflow: fast thinking profiles the series and selects a lightweight numerical forecaster to produce a forecast prior; slow deliberation retrieves contextual evidence, adaptively chooses look-back windows, and refines the prior; reflective evaluation checks temporal, contextual, and domain consistency and applies targeted corrections. The authors instantiate the workflow both in a training-free mode with off-the-shelf LLM coordinators and in a compact model trained with supervised fine-tuning and multi-turn reinforcement learning. Experiments on ETT, Wind, and EPF benchmarks compare CastFSR against statistical, deep, foundation, LLM-based, and agentic baselines, with ablations of the three workflow stages and of the SFT/RL training stages.

Significance. If validated, the proposed formulation is timely and useful: it reframes context-aware forecasting as sequential decision making over numerical experts and contextual evidence, explicitly separates numerical extrapolation from LLM-based reasoning, and supports both training-free inference and distillation into a compact model. The paper is also strong in shipping code, evaluating multiple LLM coordinators, and providing stage-level ablations. However, the central empirical attribution is currently underdetermined: the internal forecaster that visibly dominates the reported priors (Chronos-2) is not benchmarked as an independent baseline, and the comparison protocol does not state whether all baselines received the same future-known covariates. The strength of the claimed contribution therefore rests on experiments that are not yet reported.

major comments (4)
  1. [Implementation Details / Main Results (Table 1)] The fast-thinking model pool is stated to include ARIMA, DLinear, PatchTST, iTransformer, and Chronos-2, and the case study in Figure 6 shows a Chronos-2 forecast serving as the numerical prior. Chronos-2 is not included in Table 1 and is also absent from the best-fixed-model comparison in Table 4. Consequently, the reported improvements—particularly the large Wind margin (CastFSR-Zero MSE 1.596 vs. AlphaCast 2.375)—could be explained by the raw Chronos-2 prior combined with future-known covariates, rather than by the Fast–Slow–Reflect orchestration. Please report Chronos-2 alone under the same protocol and compare CastFSR against a fixed-prior variant that uses only Chronos-2 with the slow-deliberation and reflection stages removed, so that the attribution to the workflow is isolated.
  2. [Experimental Settings] The paper specifies unified look-back/horizon settings but does not state which baselines received the same future-known contextual covariates that CastFSR feeds through H_ctx in Eq. (5) and through the prompt. TimeXer, TimesFM, Sundial, and the EPF baselines differ in their native use of exogenous variables; if CastFSR had access to weather or load forecasts that the baselines did not, the reported advantage could reflect an information asymmetry rather than the Fast–Slow–Reflect design. Please state explicitly, for each baseline, which covariates were provided, or rerun the comparisons with all methods given the same covariate inputs.
  3. [Tables 1, 3, 6, 8, 9] All tables report single-run point estimates with no standard deviations, confidence intervals, or significance tests. This is particularly consequential because many reported differences are very small (e.g., Table 8 shows ETTh1 MSE 0.081 vs. 0.080 across coordinators, and Table 3 and Table 5 contain several exact ties at 0.055). Without variance estimates or repeated trials, the claim that CastFSR 'consistently outperforms' baselines is not statistically supported. Please add multiple seeds/runs, error bars, and a significance or effect-size analysis for the main comparisons and ablations.
  4. [Ablation Study of the Workflow Modules (Table 3)] The 'w/o Fast-thinking' variant removes all numerical forecasting, so it cannot distinguish the contribution of the orchestration from the contribution of the particular internal forecaster. The ablation also does not include a variant that fixes the prior to Chromos-2 while only removing slow deliberation or reflection. Without such a control, the stage-level ablations do not identify whether the gains come from the workflow or from the model pool. Please add a fixed-expert ablation (e.g., always Chronos-2 prior, with and without slow/reflective stages) and a variant that replaces Chronos-2 in the pool with a weaker expert, to better separate these factors.
minor comments (5)
  1. [Table 1 caption] The caption states that best results are bold and second-best underlined, but in the manuscript text the formatting is not visible; please ensure the typeset table clearly marks these and state the number of best/second-best counts per method.
  2. [Equation (9)] The normalized advantage A_i = (R_i - mu_R)/(sigma_R + epsilon) uses mu_R and sigma_R without defining the reference distribution; please clarify whether these are computed over the group G and how epsilon is chosen.
  3. [Figure 5] The 'w/o Contextual Cognition' variant is not defined in the experimental settings; please specify what is removed relative to CastFSR-Zero and whether the model pool and reflection stages remain unchanged.
  4. [Main Results] The sentence 'CastFSR achieves the best or second-best performance on most metrics' is not quantified; please state explicitly how many of the 20 metric–dataset cells are best or second-best for each CastFSR variant.
  5. [Prompt Template / Appendix] The prompt template refers to 'target-blind, dataset-specific rule gate' and 'coarse target-blind calibration,' but the paper does not define what information the rule gate uses or how calibration is performed; please add a precise description of these mechanisms and their inputs.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: reported gains are external benchmark comparisons; the Chronos-2-inside-the-pool issue is an attribution confound, not a circular derivation.

full rationale

CastFSR's formal chain (Eqs. 4-8) treats the numerical prior, contextual refinement, and reflective correction as a policy over observable inputs; none of these equations defines the evaluation metric, and the reported MSE/MAE values are measured against external public benchmarks. The RL objective (Eq. 9 and the accompanying reward description) is trained on trajectory rewards rather than on the test tables, so no fitted parameter is being renamed as a prediction. Self-citations (TokenCast, TimeReasoner, AlphaCast, MemCast, ConvTimeNet) appear as baselines or related-work positioning; the central claim does not depend on any of these prior results being true, and no uniqueness theorem or ansatz is imported from them. The skeptic's concern is real but is not circularity: the Implementation Details state that the model pool includes ARIMA, DLinear, PatchTST, iTransformer, and Chronos-2, while Table 1 and Table 4 do not report Chronos-2 as a standalone baseline or fixed-model comparator, and the baseline protocol does not state that every baseline received the same future-known covariates. That leaves the marginal contribution of the Fast-Slow-Reflect orchestration underdetermined, but it is a missing-control and attribution problem rather than a definitional or self-citation reduction. I cannot exhibit a step in which a reported prediction is equivalent by construction to an input or to a self-cited claim, so the circularity score is 0.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central claim rests mainly on standard ML experimental assumptions and on domain assumptions that external context variables are available and informative. The framework introduces no new physical or theoretical entities. Hand-selected design choices (look-back, learning rates, reward composition, model pool) are not fitted to the test benchmarks, but they are choices that could affect the outcome and are not fully justified.

free parameters (6)
  • look-back/horizon settings = 96/96 long-term, 168/24 short-term
    Chosen by the authors following prior benchmarks; they apply uniformly to all methods.
  • SFT learning rate = 2e-6
    Hyperparameter for training CastFSR-R1.
  • RL learning rate = 2e-7
    Hyperparameter for GRPO training.
  • RL group size G = 5
    Number of trajectories per group for GRPO.
  • maximum output length = 32768 tokens
    Inference setting for the coordinator LLM.
  • reward composition = not specified numerically
    The episode reward combines validity, numerical accuracy, and structural agreement; exact weights are not reported, so the trade-off is a hidden hand choice.
assumptions (5)
  • domain assumption The target series dynamics are influenced by both historical patterns and external contextual features.
    Foundational premise of the framework, stated in the Introduction.
  • domain assumption Future-known contextual covariates (weather forecasts, load forecasts) are available at inference time.
    The slow deliberative stage retrieves these to adjust the prior; without them the method loses its main lever.
  • domain assumption The LLM coordinator can reliably select forecasters, retrieve evidence, and reason about contextual influence.
    The entire orchestration depends on the LLM's tool-use and reasoning reliability; the paper tests this empirically but does not prove it.
  • domain assumption At least one forecaster in the pool (ARIMA, DLinear, PatchTST, iTransformer, Chronos-2) provides a useful prior for each test series.
    If the pool were uniformly poor, the prior would be poor and context adjustments would start from a bad baseline.
  • domain assumption Standard train/test splits with no target leakage into prompts.
    The prompt is target-blind, but the paper provides no explicit leakage audit beyond the validation filter described for SFT trajectories.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CastFSR: A Fast--Slow--Reflect Agentic Reasoning Framework for Context-Aware Time Series Forecasting." pith.science (2026). https://pith.science/paper/K3A6PRQ5

@misc{pith2026260803031,
  author       = {Pith},
  title        = {Pith review of: CastFSR: A Fast--Slow--Reflect Agentic Reasoning Framework for Context-Aware Time Series Forecasting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/K3A6PRQ5}},
  note         = {Machine review of arXiv:2608.03031}
}
read the original abstract

Time series forecasting is fundamental to decision-making in complex systems, where future dynamics are influenced not only by historical observations but also by evolving contextual features. Recent advances in large language models (LLMs) have extended forecasting beyond numerical extrapolation toward context-aware reasoning. However, existing approaches often lack explicit mechanisms to identify relevant contexts, reason about their impacts, and validate forecasts against temporal and domain constraints. In this work, we propose CastFSR, an agentic framework that formulates context-aware forecasting as a Fast--Slow--Reflect workflow. In fast thinking, CastFSR profiles observations and selects lightweight forecasters to construct a data-driven forecast prior. In slow deliberation, it retrieves contextual evidence, adaptively determines informative look-back windows, and reasons about how contexts reshape future dynamics. In reflection, it iteratively refines forecasts to ensure temporal, contextual, and domain consistency. CastFSR supports both training-free inference with off-the-shelf LLMs and efficient deployment through a two-stage SFT and reinforcement learning strategy that transfers its orchestration capability to compact LLMs. Extensive experiments on public datasets demonstrate that CastFSR consistently outperforms representative baselines. Our code is available at https://github.com/Xiaoyu-Tao/CastFSR.

Figures

Figures reproduced from arXiv: 2608.03031 by the authors.

Figure 1
Figure 1. Illustration of the CastFSR, which integrates nu [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of the CastFSR framework for agentic time series forecasting. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Feature-conditioned model selection patterns. Model-selection frequencies (%) conditioned on diagnostic features. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Analysis of adaptive slow deliberative reasoning, [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Analysis of context utilization strategies. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Case study of CastFSR for context-aware TSF. [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Case study of adaptive look-back window selection in CastFSR. The upper examples show successful cases where [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

154 extracted references · 33 canonical work pages

  1. [1]

    Communication, Simulation, and Intelligent Agents: Implications of Personal Intelligent Machines for Medical Education

    Clancey, William J. Communication, Simulation, and Intelligent Agents: Implications of Personal Intelligent Machines for Medical Education. Proceedings of the Eighth International Joint Conference on Artificial Intelligence (IJCAI-83)

  2. [2]

    Classification Problem Solving

    Clancey, William J. Classification Problem Solving. Proceedings of the Fourth National Conference on Artificial Intelligence

  3. [3]

    , title =

    Robinson, Arthur L. , title =. 1980 , doi =. https://science.sciencemag.org/content/208/4447/1019.full.pdf , journal =

    1980

  4. [4]

    New Ways to Make Microcircuits Smaller---Duplicate Entry

    Robinson, Arthur L. New Ways to Make Microcircuits Smaller---Duplicate Entry. Science

  5. [5]

    Clancey and Glenn Rennels , abstract =

    Diane Warner Hasling and William J. Clancey and Glenn Rennels , abstract =. Strategic explanations for a diagnostic consultation system , journal =. 1984 , issn =. doi:https://doi.org/10.1016/S0020-7373(84)80003-6 , url =

    doi:10.1016/s0020-7373(84)80003-6 1984

  6. [6]

    and Rennels, Glenn R

    Hasling, Diane Warner and Clancey, William J. and Rennels, Glenn R. and Test, Thomas. Strategic Explanations in Consultation---Duplicate. The International Journal of Man-Machine Studies

  7. [7]

    Poligon: A System for Parallel Problem Solving

    Rice, James. Poligon: A System for Parallel Problem Solving

  8. [8]

    Transfer of Rule-Based Expertise through a Tutorial Dialogue

    Clancey, William J. Transfer of Rule-Based Expertise through a Tutorial Dialogue

Show all 154 references
  1. [9]

    The Engineering of Qualitative Models

    Clancey, William J. The Engineering of Qualitative Models

  2. [10]

    2023 , eprint=

    Attention Is All You Need , author=. 2023 , eprint=

    2023

  3. [11]

    Pluto: The 'Other' Red Planet

    NASA. Pluto: The 'Other' Red Planet

  4. [13]

    R.; Marcotte, \'E .; Zantedeschi, V.; Subramanian, J.; Riachi, R.; Requeima, J.; Lacoste, A.; Rish, I.; Chapados, N.; et al

    Ashok, A.; Williams, A. R.; Marcotte, \'E .; Zantedeschi, V.; Subramanian, J.; Riachi, R.; Requeima, J.; Lacoste, A.; Rish, I.; Chapados, N.; et al. 2024. Context is key: a benchmark for forecasting with essential textual information. In NeurIPS Workshop on Time Series in the ...

    2024

  5. [15]

    N.; Hoos, H.; B \"a ck, T.; and Olhofer, M

    Baratchi, M.; Wang, C.; Limmer, S.; Van Rijn, J. N.; Hoos, H.; B \"a ck, T.; and Olhofer, M. 2024. Automated machine learning: past, present and future. Artificial intelligence review, 57(5): 122

    2024

  6. [16]

    Box, G. E. P.; Jenkins, G. M.; Reinsel, G. C.; and Ljung, G. M. 2015. Time Series Analysis: Forecasting and Control. Wiley, 5 edition

    2015

  7. [18]

    Cheng, M.; Liu, Z.; Tao, X.; Liu, Q.; Zhang, J.; Pan, T.; Zhang, S.; He, P.; Zhang, X.; Wang, D.; et al. 2025 a . A Comprehensive Survey of Time Series Forecasting: Concepts, Challenges, and Future Directions. Authorea Preprints

    2025

  8. [19]

    Cheng, M.; Wang, J.; Wang, D.; Tao, X.; Liu, Q.; and Chen, E. 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

    2026

  9. [20]

    Cheng, M.; Yang, J.; Pan, T.; Liu, Q.; Li, Z.; and Wang, S. 2025 b . Convtimenet: A deep hierarchical fully convolutional model for multivariate time series analysis. In Companion Proceedings of the ACM on Web Conference 2025, 171--180

    2025

  10. [21]

    Das, A.; Kong, W.; Sen, R.; and Zhou, Y. 2024. A decoder-only foundation model for time-series forecasting. In Forty-first International Conference on Machine Learning

    2024

  11. [22]

    Deng, J.; Chen, X.; Jiang, R.; Song, X.; and Tsang, I. W. 2022. A multi-view multi-task learning framework for multi-variate time series forecasting. IEEE Transactions on Knowledge and Data Engineering, 35(8): 7665--7680

    2022

  12. [23]

    Gao, C.; Lan, X.; Li, N.; Yuan, Y.; Ding, J.; Zhou, Z.; Xu, F.; and Li, Y. 2024. Large language models empowered agent-based modeling and simulation: A survey and perspectives. Humanities and Social Sciences Communications, 11(1): 1--24

    2024

  13. [24]

    Guo, S.; Lin, Y.; Feng, N.; Song, C.; and Wan, H. 2019. Attention based spatial-temporal graph convolutional networks for traffic flow forecasting. In Proceedings of the AAAI conference on artificial intelligence, volume 33, 922--929

    2019

  14. [25]

    Holland, J. H. 1992. Complex adaptive systems. Daedalus, 121(1): 17--30

    1992

  15. [26]

    J.; and Khandakar, Y

    Hyndman, R. J.; and Khandakar, Y. 2008. Automatic time series forecasting: the forecast package for R. Journal of statistical software, 27: 1--22

    2008

  16. [27]

    Ji, J.; Wang, J.; Huang, C.; Wu, J.; Xu, B.; Wu, Z.; Zhang, J.; and Zheng, Y. 2023. Spatio-temporal self-supervised learning for traffic flow prediction. In Proceedings of the AAAI conference on artificial intelligence, volume 37, 4356--4364

    2023

  17. [28]

    Jia, F.; Wang, K.; Zheng, Y.; Cao, D.; and Liu, Y. 2024. Gpt4mts: Prompt-based large language model for multimodal time-series forecasting. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, 23343--23351

    2024

  18. [29]

    Jin, G.; Liang, Y.; Fang, Y.; Shao, Z.; Huang, J.; Zhang, J.; and Zheng, Y. 2023. Spatio-temporal graph neural networks for predictive learning in urban computing: A survey. IEEE transactions on knowledge and data engineering, 36(10): 5388--5408

    2023

  19. [30]

    Y.; Shi, X.; Chen, P.-Y.; Liang, Y.; Li, Y.-F.; Pan, S.; et al

    Jin, M.; Wang, S.; Ma, L.; Chu, Z.; Zhang, J. Y.; Shi, X.; Chen, P.-Y.; Liang, Y.; Li, Y.-F.; Pan, S.; et al. 2024. Time-LLM: Time Series Forecasting by Reprogramming Large Language Models. In The Twelfth International Conference on Learning Representations

    2024

  20. [31]

    Lago, J.; Marcjasz, G.; De Schutter, B.; and Weron, R. 2021. Forecasting day-ahead electricity prices: A review of state-of-the-art algorithms, best practices and an open-access benchmark. Applied Energy, 293: 116983

    2021

  21. [32]

    Li, Y.; Lu, X.; Wang, Y.; and Dou, D. 2022. Generative time series forecasting with diffusion, denoise, and disentanglement. Advances in Neural Information Processing Systems, 35: 23009--23022

    2022

  22. [33]

    Liu, C.; Xu, Q.; Miao, H.; Yang, S.; Zhang, L.; Long, C.; Li, Z.; and Zhao, R. 2025 a . Timecma: Towards llm-empowered multivariate time series forecasting via cross-modality alignment. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 39, 18780--18788

    2025

  23. [34]

    Liu, Y.; Hu, T.; Zhang, H.; Wu, H.; Wang, S.; Ma, L.; and Long, M. 2024. iTransformer: Inverted Transformers Are Effective for Time Series Forecasting. In The Twelfth International Conference on Learning Representations

    2024

  24. [35]

    Liu, Y.; Qin, G.; Shi, Z.; Chen, Z.; Yang, C.; Huang, X.; Wang, J.; and Long, M. 2025 b . Sundial: A Family of Highly Capable Time Series Foundation Models. In International Conference on Machine Learning, 39295--39317. PMLR

    2025

  25. [36]

    Lu, J.; Chen, P.; Guo, C.; Shu, Y.; Wang, M.; and Yang, B. 2026. Towards non-stationary time series forecasting with temporal stabilization and frequency differencing. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 40, 24070--24078

    2026

  26. [38]

    P.; Medeiros, M

    Masini, R. P.; Medeiros, M. C.; and Mendes, E. F. 2023. Machine learning advances for time series forecasting. Journal of economic surveys, 37(1): 76--111

    2023

  27. [39]

    Nguyen, N.; Sinthong, P.; and Kalagnanam, J

    Nie, Y.; H. Nguyen, N.; Sinthong, P.; and Kalagnanam, J. 2023. A Time Series is Worth 64 Words: Long-term Forecasting with Transformers. In International Conference on Learning Representations

    2023

  28. [40]

    H.; Dayama, P.; Sindhgatta, R.; Mohapatra, P.; et al

    Palaskar, S.; Ekambaram, V.; Jati, A.; Gantayat, N.; Saha, A.; Nagar, S.; Nguyen, N. H.; Dayama, P.; Sindhgatta, R.; Mohapatra, P.; et al. 2024. Automixer for improved multivariate time-series forecasting on business and it observability data. In Proceedings of the AAAI confer...

    2024

  29. [41]

    Pan, Z.; Jiang, Y.; Garg, S.; Schneider, A.; Nevmyvaka, Y.; and Song, D. 2024. S 2 IP-LLM: Semantic Space Informed Prompt Learning with LLM for Time Series Forecasting. In Forty-first International Conference on Machine Learning

    2024

  30. [42]

    Shao, Z.; Wang, F.; Xu, Y.; Wei, W.; Yu, C.; Zhang, Z.; Yao, D.; Sun, T.; Jin, G.; Cao, X.; et al. 2024 a . Exploring progress in multivariate time series forecasting: Comprehensive benchmarking and heterogeneity analysis. IEEE Transactions on Knowledge and Data Engineering, 3...

    2024

  31. [44]

    Shi, X.; Wang, S.; Nie, Y.; Li, D.; Ye, Z.; Wen, Q.; and Jin, M. 2025. Time-MoE: Billion-Scale Time Series Foundation Models with Mixture of Experts. In The Thirteenth International Conference on Learning Representations

    2025

  32. [47]

    J.; and Letham, B

    Taylor, S. J.; and Letham, B. 2018. Forecasting at scale. The American Statistician, 72(1): 37--45

    2018

  33. [48]

    Wang, F.; Zhang, Z.; Zhang, X.; Wu, Z.; Mo, T.; Lu, Q.; Wang, W.; Li, R.; Xu, J.; Tang, X.; et al. 2025. A comprehensive survey of small language models in the era of large language models: Techniques, enhancements, applications, collaboration with llms, and trustworthiness. A...

    2025

  34. [49]

    Wang, Y.; Smola, A.; Maddix, D.; Gasthaus, J.; Foster, D.; and Januschowski, T. 2019. Deep factors for forecasting. In International conference on machine learning, 6607--6617. PMLR

    2019

  35. [50]

    Wang, Y.; Wu, H.; Dong, J.; Qin, G.; Zhang, H.; Liu, Y.; Qiu, Y.; Wang, J.; and Long, M. 2024. Timexer: Empowering transformers for time series forecasting with exogenous variables. Advances in Neural Information Processing Systems, 37: 469--498

    2024

  36. [51]

    V.; Zhou, D.; et al

    Wei, J.; Wang, X.; Schuurmans, D.; Bosma, M.; Xia, F.; Chi, E.; Le, Q. V.; Zhou, D.; et al. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35: 24824--24837

    2022

  37. [52]

    Winters, P. R. 1960. Forecasting sales by exponentially weighted moving averages. Management science, 6(3): 324--342

    1960

  38. [53]

    Woo, G.; Liu, C.; Kumar, A.; Xiong, C.; Savarese, S.; and Sahoo, D. 2024. Unified training of universal time series forecasting transformers. In Forty-first International Conference on Machine Learning

    2024

  39. [54]

    Wu, H.; Xu, J.; Wang, J.; and Long, M. 2021. Autoformer: Decomposition transformers with auto-correlation for long-term series forecasting. Advances in neural information processing systems, 34: 22419--22430

    2021

  40. [57]

    Xue, H.; and Salim, F. D. 2023. Promptcast: A new prompt-based learning paradigm for time series forecasting. IEEE Transactions on Knowledge and Data Engineering, 36(11): 6851--6864

    2023

  41. [58]

    R.; and Cao, Y

    Yao, S.; Zhao, J.; Yu, D.; Du, N.; Shafran, I.; Narasimhan, K. R.; and Cao, Y. 2022. React: Synergizing reasoning and acting in language models. In The eleventh international conference on learning representations

    2022

  42. [59]

    Zeng, A.; Chen, M.; Zhang, L.; and Xu, Q. 2023. Are transformers effective for time series forecasting? In Proceedings of the AAAI conference on artificial intelligence, volume 37, 11121--11128

    2023

  43. [60]

    Zhang, W.; Liu, H.; Liu, Y.; Zhou, J.; and Xiong, H. 2020. Semi-supervised hierarchical recurrent graph neural network for city-wide parking availability prediction. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 34, 1186--1193

    2020

  44. [63]

    Zhou, H.; Zhang, S.; Peng, J.; Zhang, S.; Li, J.; Xiong, H.; and Zhang, W. 2021. Informer: Beyond efficient transformer for long sequence time-series forecasting. In Proceedings of the AAAI conference on artificial intelligence, volume 35, 11106--11115

    2021

  45. [64]

    Zhou, T.; Niu, P.; Sun, L.; Jin, R.; et al. 2023. One fits all: Power general time series analysis by pretrained lm. Advances in neural information processing systems, 36: 43322--43355

    2023

  46. [65]

    Authorea Preprints , year=

    A Comprehensive Survey of Time Series Forecasting: Concepts, Challenges, and Future Directions , author=. Authorea Preprints , year=

  47. [66]

    arXiv preprint arXiv:2403.20150 , year=

    Tfb: Towards comprehensive and fair benchmarking of time series forecasting methods , author=. arXiv preprint arXiv:2403.20150 , year=

    arXiv

  48. [67]

    ACM Transactions on Information Systems (TOIS) , volume=

    Temporal relational ranking for stock prediction , author=. ACM Transactions on Information Systems (TOIS) , volume=. 2019 , publisher=

    2019

  49. [68]

    International Conference on Machine Learning , volumn=

    TimeDART: A Diffusion Autoregressive Transformer for Self-Supervised Time Series Representation , author=. International Conference on Machine Learning , volumn=

  50. [69]

    Journal of statistical software , volume=

    Automatic time series forecasting: the forecast package for R , author=. Journal of statistical software , volume=

  51. [70]

    Management science , volume=

    Forecasting sales by exponentially weighted moving averages , author=. Management science , volume=. 1960 , publisher=

    1960

  52. [71]

    Forty-second International Conference on Machine Learning , year=

    Temporal Query Network for Efficient Multivariate Time Series Forecasting , author=. Forty-second International Conference on Machine Learning , year=

  53. [72]

    Forty-second International Conference on Machine Learning , year=

    Hi-Patch: Hierarchical Patch GNN for Irregular Multivariate Time Series , author=. Forty-second International Conference on Machine Learning , year=

  54. [73]

    Companion Proceedings of the ACM on Web Conference 2025 , pages=

    Convtimenet: A deep hierarchical fully convolutional model for multivariate time series analysis , author=. Companion Proceedings of the ACM on Web Conference 2025 , pages=

    2025

  55. [74]

    Proceedings of the AAAI conference on artificial intelligence , volume=

    Are transformers effective for time series forecasting? , author=. Proceedings of the AAAI conference on artificial intelligence , volume=

  56. [75]

    Forty-Second International Conference on Machine Learning (ICML 2025) , year=

    Time-VLM: Exploring Multimodal Vision-Language Models for Augmented Time Series Forecasting , author=. Forty-Second International Conference on Machine Learning (ICML 2025) , year=

    2025

  57. [76]

    Soft Computing , volume=

    Temporal convolutional neural (TCN) network for an effective weather forecasting using time-series data from the local weather station , author=. Soft Computing , volume=. 2020 , publisher=

    2020

  58. [77]

    Forty-second International Conference on Machine Learning , year=

    Towards a General Time Series Forecasting Model with Unified Representation and Adaptive Transfer , author=. Forty-second International Conference on Machine Learning , year=

  59. [78]

    The 41st international ACM SIGIR conference on research & development in information retrieval , pages=

    Modeling long-and short-term temporal patterns with deep neural networks , author=. The 41st international ACM SIGIR conference on research & development in information retrieval , pages=

  60. [79]

    Proceedings of the Nineteenth ACM International Conference on Web Search and Data Mining , pages=

    Can slow-thinking llms reason over time? empirical studies in time series forecasting , author=. Proceedings of the Nineteenth ACM International Conference on Web Search and Data Mining , pages=

  61. [80]

    arXiv preprint arXiv:2509.11575 , year=

    A survey of reasoning and agentic systems in time series with large language models , author=. arXiv preprint arXiv:2509.11575 , year=

    arXiv

  62. [81]

    arXiv preprint arXiv:2512.10696 , year=

    Remember me, refine me: A dynamic procedural memory framework for experience-driven agent evolution , author=. arXiv preprint arXiv:2512.10696 , year=

    arXiv

  63. [82]

    International conference on machine learning , pages=

    Deep factors for forecasting , author=. International conference on machine learning , pages=. 2019 , organization=

    2019

  64. [83]

    The Thirteenth International Conference on Learning Representations , year=

    Time-MoE: Billion-Scale Time Series Foundation Models with Mixture of Experts , author=. The Thirteenth International Conference on Learning Representations , year=

  65. [84]

    Forty-second International Conference on Machine Learning , year=

    LangTime: A Language-Guided Unified Model for Time Series Forecasting with Proximal Policy Optimization , author=. Forty-second International Conference on Machine Learning , year=

  66. [85]

    The Twelfth International Conference on Learning Representations , year=

    Time-LLM: Time Series Forecasting by Reprogramming Large Language Models , author=. The Twelfth International Conference on Learning Representations , year=

  67. [86]

    arXiv preprint arXiv:2506.11040 , year=

    Large Language models for Time Series Analysis: Techniques, Applications, and Challenges , author=. arXiv preprint arXiv:2506.11040 , year=

    arXiv

  68. [87]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Timecma: Towards llm-empowered multivariate time series forecasting via cross-modality alignment , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  69. [88]

    IEEE Transactions on Knowledge and Data Engineering , volume=

    Promptcast: A new prompt-based learning paradigm for time series forecasting , author=. IEEE Transactions on Knowledge and Data Engineering , volume=. 2023 , publisher=

    2023

  70. [89]

    arXiv preprint arXiv:2402.16132 , year=

    LSTPrompt: Large Language Models as Zero-Shot Time Series Forecasters by Long-Short-Term Prompting , author=. arXiv preprint arXiv:2402.16132 , year=

    arXiv

  71. [90]

    arXiv preprint arXiv:2602.01776 , year=

    Position: Beyond Model-Centric Prediction--Agentic Time Series Forecasting , author=. arXiv preprint arXiv:2602.01776 , year=

  72. [91]

    Forty-first International Conference on Machine Learning , year=

    Unified training of universal time series forecasting transformers , author=. Forty-first International Conference on Machine Learning , year=

  73. [92]

    arXiv preprint arXiv:2508.06471 , year=

    Glm-4.5: Agentic, reasoning, and coding (arc) foundation models , author=. arXiv preprint arXiv:2508.06471 , year=

    arXiv

  74. [93]

    arXiv preprint arXiv:2601.13653 , year=

    TimeART: Towards Agentic Time Series Reasoning via Tool-Augmentation , author=. arXiv preprint arXiv:2601.13653 , year=

  75. [94]

    arXiv preprint arXiv:2506.10630 , year=

    Time Series Forecasting as Reasoning: A Slow-Thinking Approach with Reinforced LLMs , author=. arXiv preprint arXiv:2506.10630 , year=

    arXiv

  76. [95]

    Proceedings of the 42nd International Conference on Machine Learning , series=

    TimePro: Efficient Multivariate Long-term Time Series Forecasting with Variable- and Time-Aware Hyper-state , author=. Proceedings of the 42nd International Conference on Machine Learning , series=. 2025 , publisher=

    2025

  77. [96]

    arXiv preprint arXiv:2601.10029 , year=

    PaperScout: An Autonomous Agent for Academic Paper Search with Process-Aware Sequence-Level Policy Optimization , author=. arXiv preprint arXiv:2601.10029 , year=

  78. [97]

    arXiv preprint arXiv:2504.20073 , year=

    Ragen: Understanding self-evolution in llm agents via multi-turn reinforcement learning , author=. arXiv preprint arXiv:2504.20073 , year=

    arXiv

  79. [98]

    Proceedings of the Nineteenth ACM International Conference on Web Search and Data Mining , pages=

    Tablemind: An autonomous programmatic agent for tool-augmented table reasoning , author=. Proceedings of the Nineteenth ACM International Conference on Web Search and Data Mining , pages=

  80. [99]

    arXiv preprint arXiv:2511.14460 , year=

    Agent-r1: Training powerful llm agents with end-to-end reinforcement learning , author=. arXiv preprint arXiv:2511.14460 , year=

    arXiv

  81. [100]

    arXiv preprint arXiv:2510.01538 , year=

    Timeseriesscientist: A general-purpose ai agent for time series analysis , author=. arXiv preprint arXiv:2510.01538 , year=

  82. [101]

    2015 , edition=

    Time Series Analysis: Forecasting and Control , author=. 2015 , edition=

    2015

  83. [102]

    Journal of economic surveys , volume=

    Machine learning advances for time series forecasting , author=. Journal of economic surveys , volume=. 2023 , publisher=

    2023

  84. [103]

    IEEE Transactions on Knowledge and Data Engineering , volume=

    The capacity and robustness trade-off: Revisiting the channel independent strategy for multivariate time series forecasting , author=. IEEE Transactions on Knowledge and Data Engineering , volume=. 2024 , publisher=

    2024

  85. [104]

    Feuerriegel, S., Hartmann, J., Janiesch, C., and Zschech, P , pages=

    Introduction to reinforcement learning , author=. Feuerriegel, S., Hartmann, J., Janiesch, C., and Zschech, P , pages=

  86. [105]

    arXiv preprint arXiv:2403.07815 , year=

    Chronos: Learning the language of time series , author=. arXiv preprint arXiv:2403.07815 , year=

    arXiv

  87. [106]

    Advances in neural information processing systems , volume=

    Chain-of-thought prompting elicits reasoning in large language models , author=. Advances in neural information processing systems , volume=

  88. [107]

    arXiv preprint arXiv:2508.09191 , year=

    From values to tokens: An llm-driven framework for context-aware time series forecasting via symbolic discretization , author=. arXiv preprint arXiv:2508.09191 , year=

    arXiv

  89. [108]

    arXiv preprint arXiv:2511.08947 , year=

    AlphaCast: A Human Wisdom-LLM Intelligence Co-Reasoning Framework for Interactive Time Series Forecasting , author=. arXiv preprint arXiv:2511.08947 , year=

    arXiv

  90. [109]

    NeurIPS Workshop on Time Series in the Age of Large Models , year=

    Context is key: a benchmark for forecasting with essential textual information , author=. NeurIPS Workshop on Time Series in the Age of Large Models , year=

  91. [110]

    Advances in neural information processing systems , volume=

    Autoformer: Decomposition transformers with auto-correlation for long-term series forecasting , author=. Advances in neural information processing systems , volume=

  92. [111]

    Advances in Neural Information Processing Systems , volume=

    Adaptive normalization for non-stationary time series forecasting: A temporal slice perspective , author=. Advances in Neural Information Processing Systems , volume=

  93. [112]

    International Journal of Machine Learning and Cybernetics , pages=

    Deep learning for time series forecasting: a survey , author=. International Journal of Machine Learning and Cybernetics , pages=. 2025 , publisher=

    2025

  94. [113]

    arXiv preprint arXiv:2303.18223 , volume=

    A survey of large language models , author=. arXiv preprint arXiv:2303.18223 , volume=

    arXiv

  95. [114]

    Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=

    Contextual Experience Replay for Self-Improvement of Language Agents , author=. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=

  96. [115]

    Applied Energy , volume=

    Forecasting day-ahead electricity prices: A review of state-of-the-art algorithms, best practices and an open-access benchmark , author=. Applied Energy , volume=. 2021 , publisher=

    2021

  97. [116]

    Proceedings of the AAAI conference on artificial intelligence , volume=

    Informer: Beyond efficient transformer for long sequence time-series forecasting , author=. Proceedings of the AAAI conference on artificial intelligence , volume=

  98. [117]

    2006 , institution=

    US MOPEX data set , author=. 2006 , institution=

    2006

  99. [118]

    International Conference on Learning Representations , year =

    A Time Series is Worth 64 Words: Long-term Forecasting with Transformers , author =. International Conference on Learning Representations , year =

  100. [119]

    The American Statistician , volume=

    Forecasting at scale , author=. The American Statistician , volume=. 2018 , publisher=

    2018

  101. [120]

    The Twelfth International Conference on Learning Representations , year=

    iTransformer: Inverted Transformers Are Effective for Time Series Forecasting , author=. The Twelfth International Conference on Learning Representations , year=

  102. [121]

    Advances in Neural Information Processing Systems , volume=

    Timexer: Empowering transformers for time series forecasting with exogenous variables , author=. Advances in Neural Information Processing Systems , volume=

  103. [122]

    Advances in Neural Information Processing Systems , volume=

    Large language models are zero-shot time series forecasters , author=. Advances in Neural Information Processing Systems , volume=

  104. [123]

    arXiv preprint arXiv:2510.15821 , year=

    Chronos-2: From univariate to universal forecasting , author=. arXiv preprint arXiv:2510.15821 , year=

    arXiv

  105. [124]

    2025 , howpublished =

    2025 iFLYTEK Renewable Power Forecasting Challenge (Wind & Solar) , author =. 2025 , howpublished =

    2025

  106. [125]

    Publications of the Astronomical Society of the Pacific , volume=

    Point-Source Extraction with MOPEX , author=. Publications of the Astronomical Society of the Pacific , volume=. 2005 , publisher=

    2005

  107. [126]

    arXiv preprint arXiv:2501.12948 , year=

    Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning , author=. arXiv preprint arXiv:2501.12948 , year=

    arXiv

  108. [127]

    arXiv preprint arXiv:2411.08249 , year=

    Retrieval augmented time series forecasting , author=. arXiv preprint arXiv:2411.08249 , year=

    arXiv

  109. [128]

    ICASSP 2025-2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , pages=

    Timerag: Boosting llm time series forecasting via retrieval-augmented generation , author=. ICASSP 2025-2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , pages=. 2025 , organization=

    2025

  110. [129]

    Proceedings of the ACM on Web Conference 2025 , pages=

    Exploiting Language Power for Time Series Forecasting with Exogenous Variables , author=. Proceedings of the ACM on Web Conference 2025 , pages=

    2025

  111. [130]

    International Conference on Machine Learning , year=

    Multi-Patch Prediction: Adapting LLMs for Time Series Representation Learning , author=. International Conference on Machine Learning , year=

  112. [131]

    Pan, Zijie and Jiang, Yushan and Garg, Sahil and Schneider, Anderson and Nevmyvaka, Yuriy and Song, Dongjin , booktitle=. S

  113. [132]

    Forty-first International Conference on Machine Learning , year=

    Irregular Multivariate Time Series Forecasting: A Transformable Patching Graph Neural Networks Approach , author=. Forty-first International Conference on Machine Learning , year=

  114. [133]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Calf: Aligning llms for time series forecasting via cross-modal fine-tuning , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  115. [134]

    arXiv preprint arXiv:2303.08774 , year=

    Gpt-4 technical report , author=. arXiv preprint arXiv:2303.08774 , year=

    arXiv

  116. [135]

    Advances in neural information processing systems , volume=

    Non-stationary transformers: Exploring the stationarity in time series forecasting , author=. Advances in neural information processing systems , volume=

  117. [136]

    The Eleventh International Conference on Learning Representations , year=

    TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis , author=. The Eleventh International Conference on Learning Representations , year=

  118. [137]

    Forty-second International Conference on Machine Learning , year=

    TimeBase: The Power of Minimalism in Efficient Long-term Time Series Forecasting , author=. Forty-second International Conference on Machine Learning , year=

  119. [138]

    The Thirty-ninth Annual Conference on Neural Information Processing Systems , year=

    TimeXL: Explainable Multi-modal Time Series Prediction with LLM-in-the-Loop , author=. The Thirty-ninth Annual Conference on Neural Information Processing Systems , year=

  120. [139]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Gpt4mts: Prompt-based large language model for multimodal time-series forecasting , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  121. [140]

    ACM SIGKDD Explorations Newsletter , volume=

    Time series forecasting with llms: Understanding and enhancing model capabilities , author=. ACM SIGKDD Explorations Newsletter , volume=. 2025 , publisher=

    2025

  122. [141]

    arXiv preprint arXiv:2601.03267 , year=

    Openai gpt-5 system card , author=. arXiv preprint arXiv:2601.03267 , year=

    arXiv

  123. [142]

    arXiv preprint arXiv:2409.11376 , year=

    Towards time series reasoning with llms , author=. arXiv preprint arXiv:2409.11376 , year=

    arXiv

  124. [143]

    Advances in Neural Information Processing Systems , volume=

    Generative time series forecasting with diffusion, denoise, and disentanglement , author=. Advances in Neural Information Processing Systems , volume=

  125. [144]

    Advances in neural information processing systems , volume=

    One fits all: Power general time series analysis by pretrained lm , author=. Advances in neural information processing systems , volume=

  126. [145]

    The eleventh international conference on learning representations , year=

    React: Synergizing reasoning and acting in language models , author=. The eleventh international conference on learning representations , year=

  127. [146]

    Artificial intelligence review , volume=

    Automated machine learning: past, present and future , author=. Artificial intelligence review , volume=. 2024 , publisher=

    2024

  128. [147]

    arXiv preprint arXiv:2402.03300 , year=

    Deepseekmath: Pushing the limits of mathematical reasoning in open language models , author=. arXiv preprint arXiv:2402.03300 , year=

    arXiv

  129. [148]

    International Conference on Machine Learning , pages=

    Sundial: A Family of Highly Capable Time Series Foundation Models , author=. International Conference on Machine Learning , pages=. 2025 , organization=

    2025

  130. [149]

    Forty-first International Conference on Machine Learning , year=

    A decoder-only foundation model for time-series forecasting , author=. Forty-first International Conference on Machine Learning , year=

  131. [150]

    arXiv preprint arXiv:2309.16609 , year=

    Qwen technical report , author=. arXiv preprint arXiv:2309.16609 , year=

    arXiv

  132. [151]

    2011 , publisher=

    Thinking, Fast and Slow , author=. 2011 , publisher=

    2011

  133. [152]

    arXiv preprint arXiv:2606.19348 , year=

    Deepseek-v4: Towards highly efficient million-token context intelligence , author=. arXiv preprint arXiv:2606.19348 , year=

  134. [153]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Towards non-stationary time series forecasting with temporal stabilization and frequency differencing , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  135. [154]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Semi-supervised hierarchical recurrent graph neural network for city-wide parking availability prediction , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  136. [155]

    Proceedings of the AAAI conference on artificial intelligence , volume=

    Automixer for improved multivariate time-series forecasting on business and it observability data , author=. Proceedings of the AAAI conference on artificial intelligence , volume=

  137. [156]

    2025 International Joint Conference on Neural Networks (IJCNN) , pages=

    Mts-mixers: Multivariate time series forecasting via factorized temporal and channel mixing , author=. 2025 International Joint Conference on Neural Networks (IJCNN) , pages=. 2025 , organization=

    2025

  138. [157]

    Daedalus , volume=

    Complex adaptive systems , author=. Daedalus , volume=. 1992 , publisher=

    1992

  139. [158]

    IEEE transactions on knowledge and data engineering , volume=

    Spatio-temporal graph neural networks for predictive learning in urban computing: A survey , author=. IEEE transactions on knowledge and data engineering , volume=. 2023 , publisher=

    2023

  140. [159]

    Proceedings of the AAAI conference on artificial intelligence , volume=

    Spatio-temporal self-supervised learning for traffic flow prediction , author=. Proceedings of the AAAI conference on artificial intelligence , volume=

  141. [160]

    ACM Transactions on Intelligent Systems and Technology , volume=

    A comprehensive survey of small language models in the era of large language models: Techniques, enhancements, applications, collaboration with llms, and trustworthiness , author=. ACM Transactions on Intelligent Systems and Technology , volume=. 2025 , publisher=

    2025

  142. [161]

    Humanities and Social Sciences Communications , volume=

    Large language models empowered agent-based modeling and simulation: A survey and perspectives , author=. Humanities and Social Sciences Communications , volume=. 2024 , publisher=

    2024

  143. [162]

    Proceedings of the AAAI conference on artificial intelligence , volume=

    Attention based spatial-temporal graph convolutional networks for traffic flow forecasting , author=. Proceedings of the AAAI conference on artificial intelligence , volume=

  144. [163]

    IEEE Transactions on Knowledge and Data Engineering , volume=

    A multi-view multi-task learning framework for multi-variate time series forecasting , author=. IEEE Transactions on Knowledge and Data Engineering , volume=. 2022 , publisher=

    2022

  145. [164]

    IEEE Transactions on Knowledge and Data Engineering , volume=

    Exploring progress in multivariate time series forecasting: Comprehensive benchmarking and heterogeneity analysis , author=. IEEE Transactions on Knowledge and Data Engineering , volume=. 2024 , publisher=

    2024

  146. [165]

    arXiv preprint arXiv:2602.03164 , year=

    Memcast: Memory-driven time series forecasting with experience-conditioned reasoning , author=. arXiv preprint arXiv:2602.03164 , year=

    arXiv

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.