Pith. sign in

REVIEW 5 major objections 5 minor 50 references

Efficient Test-Time Scaling for LLM-based Time Series Forecasting

T0 review · 5 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read SCALER predicts a coarse future shape first, then lets a frozen LLM refine it, beating costlier test-time scaling baselines.

desk verdict SCALER is a solid, well-ablationed architecture — a lightweight Transformer predicts a coarse future shape that a frozen LLM refines in fixed steps — and the efficiency story plus the accuracy gains are credible, but the evidence base has enough loose ends that I'd want a revision before relying on the numbers. read the letter →

arxiv 2608.08675 v1 pith:OC2YSJ4F submitted 2026-08-09 cs.LG

Xuan-May Le , Minh-Tuan Tran , Ling Luo , Uwe Aickelin , Dinh Phung , Trung Le This is my paper · ORCID
classification cs.LG
keywords timeseriesforecastingtest-timescalinglargelanguagemodelscoarse-to-finerefinementglobalshapemodelingzero-shotmulti-scaletokenizationiterative
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

This paper tries to establish that test-time scaling for LLM-based time series forecasting can be made both cheaper and more accurate by splitting the problem into two stages: a lightweight Transformer first predicts a downscaled, coarse 'shape' of the future (trend, seasonality, regime), and a frozen pretrained LLM then refines that shape in a fixed number of coarse-to-fine residual steps. The claim is that this explicit shape anchor prevents the global-shape drift that iterative refinement suffers on long horizons, removes the need for reward-model-based candidate selection, and cuts inference cost roughly 6.7–6.8 times compared with the TimeReasoner test-time scaling baseline, while matching or beating it in accuracy. If true, the practical consequence is that LLM-based forecasters can scale test-time compute predictably and stay faithful to long-range structure without paying the full cost of slow-thinking inference.

What carries the argument

The load-bearing mechanism is the normalise-then-interpolate downscaling operator, which builds a hierarchy of future views $Y_k$ at scales $T/8$, $T/4$, $T/2$, and $T$; the coarsest view defines the shape target for the lightweight Transformer $F$, and the finer views give the token-level targets for each refinement step. A shared multi-scale patch encoder $E$ maps history and every downscaled future into the same token embedding space via prototype-based cross-attention onto the LLM's word-embedding projection, so the LLM refiner $G$ never sees raw time series. The refinement loop appends each predicted token block to the growing context $Z_S$, and the final tokens are decoded back to full resolution; training minimizes a weighted sum of shape loss, per-scale refinement loss, and full-resolution reconstruction loss. The coarse shape is doing the work: it initialises the context and constrains every subsequent step, which is what the ablation 'w/o Global Forecaster' is designed to demonstrate.

What would settle it

Run SCALER on a synthetic series whose dominant periodic component has a period that is an exact multiple of the downsampling grid (for example, a 12-step sine wave with an 8x downscale), so the coarse shape aliases the oscillation away; if the final forecast flattens or drifts while a baseline without the coarse anchor — or with a 4x downscale — recovers the periodicity, the anchor is shown to be the deciding factor. Alternatively, artificially add a ramp to the predicted coarse shape at inference and check whether the final forecast tracks the corrupted trend.

Watch

Extended reading notes

Core claim

The central claim is that predictability of the global future shape and refinement of local detail are separable tasks, and that a frozen LLM can be used as a fixed-step refiner conditioned on a coarse shape plus history tokens. SCALER encodes the historical series once into token blocks, uses a small Transformer to forecast a downscaled future, and then runs exactly K refinement steps in which the LLM predicts a finer token block that is appended to the context. Because the coarse shape anchors the whole trajectory, the LLM only adds local residuals, so trends and seasonality are preserved even on horizons of 720+ steps. The paper argues that this both outperforms direct and iterative-refinement forecasters and is far cheaper, since each refinement step processes only a compact token block and no candidate generation or reward-model scoring is needed.

Load-bearing premise

Stage I's lightweight Transformer must predict a coarse future shape accurate enough to serve as an anchor; if that shape is systematically wrong, the frozen LLM refiner can only refine around the wrong shape, and the global-shape-drift benefit vanishes.

Editorial extensions

If this is right

  • Inference cost becomes predictable and deployment-friendly: exactly K refinement steps, each over a compact token block, with no candidate sampling or reward-model selection.
  • Long-horizon accuracy degrades more gracefully than methods without a global anchor; the reported gap over TimeReasoner and TimeLLM widens for horizons beyond 720 steps.
  • The framework is largely backbone-agnostic: the paper reports competitive results with LLaMA-7B down to GPT-2 (124M), with larger backbones consistently better.
  • Zero-shot transfer between M3 and M4 is improved over the in-context-learning baseline, suggesting the coarse shape transfers across datasets without retraining.
  • Putting the global forecaster at full resolution instead of downscaled hurts accuracy as well as cost, because the forecaster's fine-grained predictions mislead the refiner.

Reading between the lines

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

  • If the anchor claim is right, the coarse forecaster is a swappable module: replacing the lightweight Transformer with a classical trend/seasonality decomposition (STL or similar) should give a near-drop-in test, since the LLM refiner only needs a faithful low-frequency guide.
  • The design implies a testable causal structure: deliberately corrupting the predicted coarse shape (e.g., adding a linear ramp) should propagate into the final forecast, shifting the whole trajectory — whereas a refiner that truly learned the series would resist the corruption.
  • The aliasing caveat in Appendix D suggests an easy stress test: on series with a strong periodic component whose period aligns with the downsampling grid, the coarse shape can miss the oscillation entirely; measuring SCALER's error on such synthetic series would delimit when the anchor helps versus hurts.
  • Fixed-step refinement could be made adaptive at test time by stopping early when successive token blocks change little, preserving the shape anchor while cutting compute further — an extension the paper leaves open.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper proposes SCALER, a two-stage coarse-to-fine framework for LLM-based time series forecasting. In Stage I, a lightweight Transformer F predicts a coarse future shape from the history. In Stage II, a frozen pretrained LLM refiner G performs a fixed K-step residual refinement, conditioned on history tokens and the predicted coarse-shape tokens, decoding the final tokens into a full-resolution forecast. The authors claim that SCALER outperforms strong baselines on long-term, short-term, and zero-shot forecasting benchmarks while reducing inference cost by roughly 6.7x-6.8x relative to the test-time-scaling baseline TimeReasoner. The paper includes extensive ablations, a code repository, and evaluations on ETT, ECL, Traffic, Weather, ILI, M3, and M4.

Significance. The central idea of anchoring iterative LLM refinement to an explicit coarse future shape is a potentially valuable contribution to efficient test-time scaling for LLM-based forecasting. If validated, the approach would offer a practical way to preserve global trend and seasonality while keeping inference cost predictable and lower than existing scaled-LLM methods. The paper's strengths include the fixed-step refinement design, the breadth of benchmarks, and the unusually thorough ablation suite (Tables 4, 7-10, 12-13), which demonstrates that each component contributes. However, the headline accuracy gains are often marginal and are reported without variance or significance testing; the load-bearing Stage-I shape anchor is not stress-tested; and the efficiency measurement is not described precisely enough to verify the main speed-up claim. The contribution is therefore promising but requires substantially stronger evidence before it can be accepted at the level claimed.

major comments (5)
  1. [§3.4, Eqs. (8)-(10), Appendix D] The central claim that SCALER preserves global shape rests on the reliability of the Stage-I coarse shape S1 predicted by F, since the LLM refiner G is conditioned on Encoding(F(X)) and can only refine around that anchor. Appendix D explicitly concedes that the normalise-then-interpolate downscaling operator risks aliasing when the downsampling grid coincides with periodic zeros, yet no spectral analysis or robustness experiment is provided to quantify the effect of F's errors on the final forecast. The ablation in Table 4 shows that removing F hurts accuracy on average, but it does not establish that the final output is robust to the magnitude or systematic direction of F's shape errors. I request an oracle-shape experiment in which F's output is replaced by the ground-truth coarse shape S1 at inference, and corrupted-shape experiments in which controlled low-frequency perturbations of increasing amplitude are added to S1, so that error propagation through the refiner can be characterized. Without these, the shape-faithfulness mechanism is not demonstrated.
  2. [§4.1, Tables 1-3] No standard deviations or significance tests are reported, and the run counts are internally inconsistent: §4.1 states results are averaged over three runs, while the captions of Tables 1-3 and the zero-shot section say five runs. Many of the headline improvements are small enough to be within plausible run-to-run noise: ECL average MSE is 0.154 for both SCALER and TimeReasoner; ETTh1 SCALER is 0.376 vs 0.381 for LVICL; Weather SCALER is 0.217 vs 0.218 for TimeReasoner, with TimeReasoner having a lower MAE (0.254 vs 0.257). Such margins cannot support the claim that SCALER 'consistently achieves the best' without variance estimates or paired significance tests, so the central accuracy claim is currently under-supported.
  3. [Table 1 caption, §4.1, Appendix E] There is a contradiction about baseline provenance. The main text in §4.1 says 'All baselines are evaluated with the same data splits, preprocessing, and metrics to ensure fair comparison,' but the Table 1 caption and Appendix E state that 'Baseline results are taken from [43]' and only TimeReasoner is rerun under the authors' setup. Copying published numbers for LVICL, AutoTimes, TimeLLM, FPT, and the other baselines rather than running them in the same codebase makes the comparison uncontrolled and undermines the claim of outperforming those methods. The authors should rerun the full baseline suite in their own pipeline, or clearly state which numbers are copied and justify why the comparison remains fair despite differences in implementation details, hyperparameters, and hardware.
  4. [§4.5, Figure 3] The claimed 6.7x-6.8x speed-up over TimeReasoner is not verifiable from the description provided. The manuscript does not report the model precision (e.g., FP16 vs BF16), the batch size used for timing, the number of tokens processed per refinement step, the GPU configuration beyond 'A100-80GB', the number of warm-up iterations, or the variance of runtime across repetitions. It is also unclear whether TimeReasoner's reward-model-based candidate selection was run with a comparable number of candidates and whether the reported ratio is per dataset or pooled. Adding a precise measurement protocol with standard errors is necessary for the efficiency claim to be reproducible.
  5. [§3.2-§3.5] The method is not fully specified in the paper. The lightweight forecaster F, the patch encoder E, the decoder D, and the exact conditioning format used to feed the LLM refiner G with (Z_I, Z_S) are described only through the losses in Eqs. (10), (17), and (19); no architecture details, token dimensions, number of layers, or LLM prompting/formatting are provided. Since this is a methods paper, the description should be complete enough to reimplement without reverse-engineering the released code. Please add a dedicated architecture subsection, including parameter counts and per-step token budgets.
minor comments (5)
  1. [Tables 2 and 3, last rows] The '1st Count' rows appear to contain formatting errors: Table 2 shows '120' where '12 0' is presumably intended, and Table 3 shows '90' where '9 0' is presumably intended. Please fix the alignment and separators.
  2. [Algorithm 2 and Eq. (12)] There are typographical errors in the context initialization: Algorithm 2, line 3 contains stray bracket characters ('Z_S ← [Encoding(Ŝ1)];') and Eq. (12) has an extra closing parenthesis ('Z_S = [S1], where S1 = Encoding(Ŝ1))').
  3. [Section 1] In the Introduction, 'In avarge inference time' should be 'On average'; please proofread the manuscript for similar typos.
  4. [§4.4] The zero-shot protocol is underspecified for the 'Others' categories: M3-Others is trained on M4-Quarterly and M4-Others on M3-Monthly, but the reasoning behind these frequency-to-category matchings is not explained, and it is unclear whether the series lengths align.
  5. [§4.1] The scale schedule is written as 'Scale ∈ T/8,,T/4,,T/2,,T' with double commas; this should be cleaned up and aligned with the {1,2,4,8} schedule used in Appendix C.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: SCALER's coarse-to-fine prediction and refinement are supervised against external data targets, and no load-bearing claim reduces by construction to a fitted input or to the authors' prior work.

full rationale

SCALER's derivation chain is self-contained and non-circular. Stage I trains a lightweight forecaster F to predict S_hat_1 against the downscaled target S_1 = Downscaling(Y; Scale_1) (Eqs. 7-10), with the coarse target defined from the ground-truth future rather than from the model's own output. Stage II trains the frozen LLM refiner G at each scale against S_k = Downscaling(Y; Scale_k) under L_refine, plus a final time-series loss L_ts against the full ground truth Y (Eqs. 13-19); at inference G conditions on history tokens and the predicted anchor but is still free to generate the refined tokens, so the final forecast is not forced to equal S_hat_1 by construction. The evaluation is against external benchmarks (ETT, ECL, Traffic, Weather, ILI, M3, M4) and external baselines, and the accuracy/efficiency comparison normalises SCALER's runtime to 1x without fitting any test quantity. The authors' prior shapelet papers [12-14, 33] are cited only as general background and are not load-bearing in the method or the uniqueness of the design. The acknowledged aliasing risk in Appendix D is a robustness caveat about the downscaling operator, not a circular reduction: the model could fail if the coarse shape is unreliable, but failure modes are empirical limitations rather than definitional equivalences. No fitted parameter is renamed as a prediction, no uniqueness theorem from the authors' prior work is invoked, and no ansatz is smuggled in via citation. Therefore the paper earns a non-finding: score 0.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The method adds a coarse-shape prediction stage, a shared multi-scale encoder, and a frozen-LLM refiner; the numerical result depends on several hand-selected hyperparameters (scale schedule, K, patch length, loss weights) tuned on the benchmark suite. The central scientific assumptions are that a small Transformer can predict a reliable coarse shape and that a frozen LLM is a useful refiner. No new physical entities are introduced.

free parameters (5)
  • Scale schedule = T/8, T/4, T/2, T (i.e., {1,2,4,8} relative steps)
    Chosen after ablation Table 9; this schedule defines the coarse-to-fine token blocks that are the core of the method, so the reported result depends on it.
  • Number of refinement steps K = 4
    Chosen after Table 9; 4 steps gave the best results and 5 steps gave no further gain.
  • Patch length = 12
    Chosen after ablation Table 8; coarser patches degrade accuracy.
  • Loss weights lambda and beta = 1 and 1
    Swept in Appendix C Table 7; the default (1,1) was best or tied on ETTh1 and ECL.
  • Learning rate and batch size = Selected from {3e-5, 5e-5, 7e-5, 9e-5} and {192, 256, 1024, 2048}
    Reported in Section 4.1 as standard validation-based model selection.
assumptions (4)
  • domain assumption A pretrained frozen LLM provides a useful generative prior for refining time-series tokens.
    Invoked in Stage II and supported only by comparison against PatchTST and iTransformer refiners (Table 12), not proven.
  • domain assumption Normalise-then-interpolate downscaling preserves the relevant low-frequency structure without harmful aliasing at the default 8x ratio.
    Used to build coarse targets in Section 3.4; Appendix D itself admits aliasing risk for high-frequency components.
  • domain assumption MSE losses on coarse shapes, refined tokens and final series are adequate training signals for forecasting quality.
    Training objective in Equation 20; does not capture probabilistic uncertainty or quantile behavior.
  • standard math Standard patch encoding and cross-attention behave as specified in Equations 1 to 3.
    The implementation relies on standard softmax attention and linear projections without formal verification.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Efficient Test-Time Scaling for LLM-based Time Series Forecasting." pith.science (2026). https://pith.science/paper/OC2YSJ4F

@misc{pith2026260808675,
  author       = {Pith},
  title        = {Pith review of: Efficient Test-Time Scaling for LLM-based Time Series Forecasting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OC2YSJ4F}},
  note         = {Machine review of arXiv:2608.08675}
}
read the original abstract

Long-term time series forecasting benefits from preserving global structure such as trends and seasonality. Recent LLM-based forecasters often improve accuracy through test-time scaling (e.g., iterative refinement), but these methods are computationally expensive and increasingly prone to global-shape mismatch as the prediction horizon extends. We propose SCALER, a coarse-to-fine forecasting framework that first employs a lightweight Transformer tailored to long-term shape modeling to predict a coarse representation of future dynamics. This predicted shape then serves as a compact guide for an LLM to perform test-time scaling via iterative coarse-to-fine residual token refinement, while processing substantially fewer tokens at each step. By guiding refinement with an explicit future-shape prediction, SCALER reduces reliance on long description prompts, and its fixed-step refinement avoids costly reward-model-based selection, further lowering computational overhead. Experimental results demonstrate that SCALER outperforms strong forecasting baselines in long-term, short-term and zero-shot forecasting while significantly reducing the inference cost associated with scaled LLM for time series forecasting. Code: https://github.com/xuanmay2701/SCALER.

Figures

Figures reproduced from arXiv: 2608.08675 by the authors.

Figure 1
Figure 1. (a) Standard: direct forecasting; may lose long-horizon [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. SCALER architecture. Given the input/history time series [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Accuracy–efficiency trade-off. MSE (y-axis, lower is better) versus relative inference time (x-axis, lower is faster) on ETTh1, ECL, and [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Effect of the depth of global forecaster. We vary the depth of [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Long-horizon robustness on ETTh1. MSE/MAE versus fore [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Representative forecast comparison. Black: ground truth; [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 28 canonical work pages

  1. [43]

    Kyung Keun Yun, Sang Won Yoon, and Daehan Won. 2023. Interpretable stock price forecasting model using genetic algorithm-machine learning regressions and best feature subset selection.Expert Systems with Applications213 (2023), 118803

    work page 2023

  2. [1]

    Abdul Fatir Ansari, Lorenzo Stella, Caner Turkmen, Xiyuan Zhang, Pedro Mercado, Huibin Shen, Oleksandr Shchur, Syama Sundar Rangapuram, Sebas- tian Pineda Arango, Shubham Kapoor, et al. 2024. Chronos: Learning the language of time series.arXiv preprint arXiv:2403.07815(2024)

    arXiv 2024

  3. [2]

    Adebiyi A Ariyo, Adewumi O Adewumi, and Charles K Ayo. 2014. Stock price prediction using the ARIMA model. In2014 UKSim-AMSS 16th international conference on computer modelling and simulation. IEEE, 106–112. KDD ’26, August 09–13, 2026, Jeju Island, Republic of Korea Xuan-May Le et al

    work page 2014

  4. [3]

    Ching Chang, Wei-Yao Wang, Wen-Chih Peng, and Tien-Fu Chen. 2025. Llm4ts: Aligning pre-trained llms as data-efficient time-series forecasters.ACM Transac- tions on Intelligent Systems and Technology16, 3 (2025), 1–20

    work page 2025

  5. [4]

    Mingyue Cheng, Jiahao Wang, Daoyu Wang, Xiaoyu Tao, Qi Liu, and Enhong Chen. 2025. Can slow-thinking llms reason over time? empirical studies in time series forecasting.arXiv preprint arXiv:2505.24511(2025)

    arXiv 2025

  6. [5]

    Abhimanyu Das, Weihao Kong, Rajat Sen, and Yichen Zhou. 2024. A decoder- only foundation model for time-series forecasting. InForty-first International Conference on Machine Learning

    2024

  7. [6]

    Jiaxin Gao, Qinglong Cao, and Yuntian Chen. 2025. Auto-regressive moving diffusion models for time series forecasting. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 39. 16727–16735

    work page 2025

  8. [7]

    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 Systems36 (2023), 19622–19635

    2023

Show all 50 references
  1. [8]

    James W Hansen, Simon J Mason, Liqiang Sun, and Arame Tall. 2011. Review of seasonal climate forecasting for agriculture in sub-Saharan Africa.Experimental agriculture47, 2 (2011), 205–240

    2011

  2. [9]

    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)

    2023 arXiv

  3. [10]

    Shruti Kaushik, Abhinav Choudhury, Pankaj Kumar Sheron, Nataraj Dasgupta, Sayee Natarajan, Larry A Pickett, and Varun Dutt. 2020. AI in healthcare: time- series forecasting using statistical, neural, and ensemble architectures.Frontiers in big data3 (2020), 4

    2020

  4. [11]

    Marcel Kollovieh, Abdul Fatir Ansari, Michael Bohlke-Schneider, Jasper Zschieg- ner, Hao Wang, and Yuyang Bernie Wang. 2023. Predict, refine, synthesize: Self-guiding diffusion models for probabilistic time series forecasting.Advances in Neural Information Processing Systems36...

    2023

  5. [12]

    Xuan-May Le, Ling Luo, Uwe Aickelin, and Minh-Tuan Tran. 2024. ShapeFormer: Shapelet Transformer for Multivariate Time Series Classification. InProceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 1484–1494

    2024

  6. [13]

    Xuan-May Le, Ling Luo, Uwe Aickelin, Minh-Tuan Tran, David Berlowitz, and Mark Howard. 2025. Ship: A shapelet-based approach for interpretable patient- ventilator asynchrony detection. InPacific-Asia Conference on Knowledge Discov- ery and Data Mining. Springer, 430–442

    2025

  7. [14]

    Xuan-May Le, Minh-Tuan Tran, and Van-Nam Huynh. 2022. Learning Percep- tual Position-Aware Shapelets for Time Series Classification. InJoint European Conference on Machine Learning and Knowledge Discovery in Databases. Springer, 53–69

    2022

  8. [15]

    Yan Li, Xinjiang Lu, Yaqing Wang, and Dejing Dou. 2022. Generative time series forecasting with diffusion, denoise, and disentanglement.Advances in Neural Information Processing Systems35 (2022), 23009–23022

    2022

  9. [16]

    Zhe Li, Shiyi Qi, Yiduo Li, and Zenglin Xu. 2023. Revisiting long-term time series forecasting: An investigation on linear mapping.arXiv preprint arXiv:2305.10721

    2023 arXiv

  10. [17]

    Chenxi Liu, Qianxiong Xu, Hao Miao, Sun Yang, Lingzheng Zhang, Cheng Long, Ziyue Li, and Rui Zhao. 2024. Timecma: Towards llm-empowered time series forecasting via cross-modality alignment.arXiv preprint arXiv:2406.01638(2024)

    2024 arXiv

  11. [18]

    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. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 39. 18915–18923

    2025

  12. [19]

    Shizhan Liu, Hang Yu, Cong Liao, Jianguo Li, Weiyao Lin, Alex X Liu, and Schahram Dustdar. 2021. Pyraformer: Low-Complexity Pyramidal Attention for Long-Range Time Series Modeling and Forecasting. InInternational Conference on Learning Representations

    2021

  13. [20]

    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)

    2023 arXiv

  14. [21]

    Yong Liu, Guo Qin, Xiangdong Huang, Jianmin Wang, and Mingsheng Long

  15. [22]

    Spyros Makridakis and Michele Hibon. 2000. The M3-Competition: results, conclusions and implications.International journal of forecasting16, 4 (2000), 451–476

    2000

  16. [23]

    Spyros Makridakis, Evangelos Spiliotis, and Vassilios Assimakopoulos. 2018. The M4 Competition: Results, findings, conclusion and way forward.International Journal of forecasting34, 4 (2018), 802–808

    2018

  17. [24]

    Luis Martín, Luis F Zarzalejo, Jesus Polo, Ana Navarro, Ruth Marchante, and Marco Cony. 2010. Prediction of global solar irradiance based on time series analysis: Application to solar thermal power plants energy production planning. Solar Energy84, 10 (2010), 1772–1781

    2010

  18. [25]

    Yasuko Matsubara, Yasushi Sakurai, Willem G Van Panhuis, and Christos Falout- sos. 2014. FUNNEL: automatic mining of spatially coevolving epidemics. In Proceedings of the 20th ACM SIGKDD international conference on Knowledge dis- covery and data mining. 105–114

    2014

  19. [26]

    Abdul Quadir Md, Sanjit Kapoor, Chris Junni AV, Arun Kumar Sivaraman, Kong Fah Tee, H Sabireen, and N Janakiraman. 2023. Novel optimization ap- proach for stock price forecasting using multi-layered sequential LSTM.Applied Soft Computing134 (2023), 109830

    2023

  20. [27]

    2012.A very British affair: Six Britons and the development of time series analysis during the 20th century

    T Mills. 2012.A very British affair: Six Britons and the development of time series analysis during the 20th century. Springer

    2012

  21. [28]

    Nguyen, Phanwadee Sinthong, and Jayant Kalagnanam

    Yuqi Nie, Nam H. Nguyen, Phanwadee Sinthong, and Jayant Kalagnanam. 2023. A Time Series is Worth 64 Words: Long-term Forecasting with Transformers. In International Conference on Learning Representations

    2023

  22. [29]

    Zheng Qian, Yan Pei, Hamidreza Zareipour, and Niya Chen. 2019. A review and discussion of decomposition-based hybrid models for wind energy forecasting applications.Applied energy235 (2019), 939–953

    2019

  23. [30]

    Kashif Rasul, Calvin Seward, Ingmar Schuster, and Roland Vollgraf. 2021. Au- toregressive denoising diffusion models for multivariate probabilistic time series forecasting. InInternational conference on machine learning. PMLR, 8857–8868

    2021

  24. [31]

    Mingtian Tan, Mike Merrill, Vinayak Gupta, Tim Althoff, and Tom Hartvigsen

  25. [32]

    Hua Tang, Chong Zhang, Mingyu Jin, Qinkai Yu, Zhenting Wang, Xiaobo Jin, Yongfeng Zhang, and Mengnan Du. 2025. Time series forecasting with llms: Understanding and enhancing model capabilities.ACM SIGKDD Explorations Newsletter26, 2 (2025), 109–118

    2025

  26. [33]

    Are language models actually useful for time series forecasting?Advances in Neural Information Processing Systems37 (2024), 60162–60191

    2024

  27. [34]

    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)

    2024 arXiv

  28. [35]

    Minh-Tuan Tran, Xuan-May Le, Van-Nam Huynh, and Sung-Eui Yoon. 2024. PISD: A linear complexity distance beats dynamic time warping on time series classification and clustering.Engineering Applications of Artificial Intelligence 138 (2024), 109222

    2024

  29. [36]

    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.Advances in Neural Information Processing Systems37 (2024), 469–498

    2024

  30. [37]

    Yihe Wang, Nan Huang, Taida Li, Yujun Yan, and Xiang Zhang. 2024. Medformer: A multi-granularity patching transformer for medical time-series classification. Advances in Neural Information Processing Systems37 (2024), 36314–36341

    2024

  31. [38]

    Haixu Wu, Yao Xu, Jialin Wang, Guodong Long, and et al. 2021. Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Fore- casting. InAdvances in Neural Information Processing Systems (NeurIPS)

    2021

  32. [39]

    Haixu Wu, Tengge Hu, Yong Liu, Hang Zhou, Jianmin Wang, and Mingsheng Long. 2022. Timesnet: Temporal 2d-variation modeling for general time series analysis.arXiv preprint arXiv:2210.02186(2022)

    2022 arXiv

  33. [40]

    Hao Xue and Flora D Salim. 2023. Promptcast: A new prompt-based learning paradigm for time series forecasting.IEEE Transactions on Knowledge and Data Engineering36, 11 (2023), 6851–6864

    2023

  34. [41]

    Haixu Wu, Hang Zhou, Mingsheng Long, and Jianmin Wang. 2023. Interpretable weather forecasting for worldwide stations with a unified deep model.Nature Machine Intelligence5, 6 (2023), 602–611

    2023

  35. [42]

    Ailing Zeng, Muxi Chen, Lei Zhang, and Qiang Xu. 2023. Are transformers effective for time series forecasting?. InProceedings of the AAAI conference on artificial intelligence, Vol. 37. 11121–11128

    2023

  36. [44]

    Yunhao Zhang and Junchi Yan. 2023. Crossformer: Transformer utilizing cross- dimension dependency for multivariate time series forecasting. InThe eleventh international conference on learning representations

    2023

  37. [45]

    Jianqi Zhang, Jingyao Wang, Wenwen Qiang, Fanjiang Xu, and Changwen Zheng

  38. [46]

    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 systems36 (2023), 43322–43355. A Datasets We evaluate on diverse real-world forecasting benchmarks covering mu...

    2023

  39. [48]

    Haoyi Zhou, Shanghang Zhang, Jieqi Peng, and et al. 2021. Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting.AAAI35, 12 (2021), 11106–11115

    2021

  40. [50]

    Following prior long-term forecasting studies, all datasets use the same preprocessing and chronological train/validation/test splits to prevent information leakage

    and M4 [23] benchmarks spanning multiple domains and fre- quencies. Following prior long-term forecasting studies, all datasets use the same preprocessing and chronological train/validation/test splits to prevent information leakage. Dataset statistics are sum- marised in Tabl...

  41. [2024]

    Advances in Neural Information Processing Systems37 (2024), 122154–122184

    Autotimes: Autoregressive time series forecasters via large language models. Advances in Neural Information Processing Systems37 (2024), 122154–122184

    2024

  42. [2026]

    Enhancing Large Language Models for Time-Series Forecasting via Vector- Injected In-Context Learning.arXiv preprint arXiv:2601.07903(2026)

    2026

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.