Pith. sign in

REVIEW 3 major objections 7 minor 41 references

M2WLLM: Multi-Modal Multi-Task Ultra-Short-term Wind Power Prediction Algorithm Based on Large Language Model

T0 review · 3 major / 7 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper claims that feeding a large language model both textual prompts and patched numerical series yields ultra-short-term wind power forecasts that beat eight baselines, including GPT4TS, on every dataset and horizon tested.

desk verdict A cleanly engineered LLM wind-forecasting architecture whose headline empirical claim is not yet trustworthy because the train/test split is not specified as chronological and no code or data are released. read the letter →

arxiv 2506.00531 v1 pith:EWUFKPWJ submitted 2025-05-31 cs.LG cs.AI

Hang Fana , Mingxuan Lib , Zuhan Zhanga , Long Chengc , Yujian Ye , Dunnan Liua This is my paper
classification cs.LGcs.AI
keywords windpowerforecastinglargelanguagemodelsmulti-modallearningtimeseriespredictionLoRAfine-tuningcross-attentionembeddingfew-shotGPT-2
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

Ultra-short-term wind power forecasts must predict output minutes to hours ahead, and this paper argues that large language models can do it better than specialized deep-learning models when given both text and numbers. M2WLLM sends a GPT-2 backbone two streams: a textual prompt describing the task, the data window, and summary statistics, and the numerical series after it has been patched and aligned with the model's pre-trained word embeddings through cross-attention. The paper reports the lowest MAE and RMSE on wind-farm data from three Chinese provinces at 15-minute, 1-hour, 2-hour, and 4-hour horizons, beating eight baselines including GPT4TS, and it degrades gracefully when trained on only 10% of the data. Ablations show that the prompt, the semantic augmenter, and LoRA fine-tuning each contribute to the gain.

What carries the argument

The load-bearing mechanism is the Semantic Augmenter, a cross-attention block that treats each patch of the normalized time series as a query and the LLM's pre-trained word-embedding matrix as keys and values. A Token Mapper first compresses the vocabulary, then multi-head cross-attention embeds each patch into a vector carrying both the numeric values and semantic associations from the pre-trained language model; these augmented tokens are concatenated with sample-specific prompts from the Prompt Embedder and fed into GPT-2, which is adapted with LoRA. The Prompt Embedder supplies a concise natural-language description of the prediction task, the data range, and the mean and variance of each feature, which the paper argues lets the LLM understand the forecasting problem rather than merely process digits.

What would settle it

Recompute Table 2 with a strictly chronological split: the first 60% of the timeline for training, the next 20% for validation, and the final 20% for testing. If M2WLLM does not still record the lowest MAE and RMSE on all three datasets at all four horizons, the central claim fails.

Watch

Extended reading notes

Core claim

The central claim is that textual context and numerical time series can be fused inside a pre-trained LLM so that the language model's linguistic knowledge improves wind power regression. Concretely, M2WLLM is reported to achieve the best MAE and RMSE on all three datasets and all four prediction horizons in Table 2, with its advantage over GPT4TS attributed to the prompt embedder and semantic augmenter rather than to the backbone alone. A second claim is few-shot competence: with 10% of the training data, M2WLLM's error is often close to, or below, the full-data error of competing methods. The paper also argues that each architectural component is necessary, because removing the prompt, replacing the semantic augmenter with a linear layer, or freezing the backbone all raise errors.

Load-bearing premise

The findings depend on the train/validation/test split being done in time order; if samples from the same timeline are shuffled, 24-hour input windows from neighboring moments can appear in both training and test sets, which would inflate the reported accuracy gains.

Editorial extensions

If this is right

  • If the reported gains hold, one LLM-based forecaster can replace separate models per wind farm and per horizon, simplifying deployment for grid operators.
  • Newly built wind farms with little history would need far less data: 10% of the training set often matches or beats full-data specialized models.
  • NWP inputs such as wind speed, pressure, and temperature matter increasingly as the forecast horizon lengthens, so ultra-short-term systems should keep these inputs even when historical power data dominates.
  • Because 4 GPT-2 layers perform nearly as well as 12, the architecture rather than raw parameter count drives the accuracy, implying cheaper LLM-based forecasting is possible.
  • The model's ability to reproduce zero-output windless periods suggests LLM semantic understanding reduces false fluctuations, which would improve ramp and shutdown detection.

Reading between the lines

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

  • Editorial extension: the paper only tests GPT-2, so whether semantic augmentation generalizes across other LLM backbones is untested and would be worth checking with a smaller, cheaper model.
  • Editorial extension: the prompt's statistics (range, mean, variance) may carry most of the information, so a testable simplification is whether the same gains appear when these statistics are appended numerically without natural-language phrasing.
  • Editorial extension: cross-attention to compressed word embeddings effectively builds a learned codebook, a mechanism that could transfer to other numerical modalities such as solar power or electricity load without redesign.
  • Editorial extension: the paper does not establish that a chronological evaluation preserves the reported lead; checking this should be the first replication step before deploying the model.
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

3 major / 7 minor

Summary. The paper proposes M2WLLM, a GPT-2-based model for ultra-short-term wind power forecasting. The model combines textual prompts with patched time series through a Prompt Embedder and a Data Embedder; the Data Embedder contains a Semantic Augmenter that performs cross-attention between patch embeddings and reduced pre-trained word embeddings. The backbone is fine-tuned with LoRA, and the output is projected to the forecast horizon for all wind-power stations simultaneously. Experiments on wind-farm data from Inner Mongolia, Gansu, and Yunnan compare MAE and RMSE against eight baselines at 1-, 4-, 8-, and 16-step horizons, with ablations of the prompt, semantic augmenter, fine-tuning, NWP information, and GPT-2 depth, plus a few-shot training-data experiment. The central claim is that M2WLLM consistently outperforms all compared methods on every dataset and horizon.

Significance. If the reported gains are valid, the paper is a useful empirical contribution showing that pre-trained LLM text embeddings can be exploited for wind-power forecasting through prompt-based task descriptions and semantic augmentation of time-series patches. The evaluation is broad, covering three datasets and four horizons, the ablations isolate several architectural choices, and the reported training and inference times provide practical context. The main comparative claim, however, rests on the integrity of the evaluation protocol, which the manuscript does not fully establish; the lack of uncertainty quantification and the unspecified data split are the central risks.

major comments (3)
  1. [Section 4.1] The paper does not state whether the 3:1:1 split of the 17,546 samples is chronological or random. Since tau_h = 96 at 15-minute resolution and the collection period is March-August 2020, the sample count is consistent with using every sliding window; under a random split, each test window's 96-step history overlaps the training windows by up to 95 time steps, so the test metrics in Table 2 would measure interpolation or memorization rather than forecasting. This directly affects the central claim in the Abstract and Section 4.4 that M2WLLM consistently outperforms all baselines. Please specify the split policy, use a chronological split or otherwise guarantee no train/test temporal overlap, and note that even a chronological split leaves overlapping test windows that reduce the effective number of independent test samples.
  2. [Section 4.4, Table 2] The comparison reports a single MAE/RMSE value per model and horizon, with no standard deviations, multiple seeds, or paired significance tests. Several margins at longer horizons are small, for example Yunnan 4-hour MAE is 7.87 for M2WLLM versus 8.20 for Adaptive-GCN and 8.28 for GPT4TS, and Inner Mongolia 2-hour RMSE is 12.21 versus 12.89 for CNN-BiLSTM-Att. Without repeated runs or statistical tests, the wording 'consistently outperforms' is not supported by the tables alone; the daily RMSE distributions in Figure 6 could be made quantitative with paired tests over test windows or a reported distribution over seeds.
  3. [Section 4.4, Figure 8] The few-shot experiment reports single MAE curves without stating how the reduced training subsets are drawn (for example, random subsample of the training set), whether each model is retrained for every percentage with the same hyperparameters, or how many repetitions are used. The strong claims that M2WLLM with 10% of the data beats other models trained on 100% (Yunnan 15-minute MAE 3.76 versus Informer 6.00) therefore lack the same statistical support as the main comparison. Please specify the protocol and report variance across repetitions.
minor comments (7)
  1. [Section 3.3.2, Eq. (4)] The symbol dLLM s in the definitions of EM, W_i^K, and W_i^V is never defined; if it is a typo for dLLM, please correct it, and state the exact dimensions of all linear projections so that the architecture is unambiguous.
  2. [Section 2.1, Eqs. (2) and Section 4.2, Eqs. (5)-(6)] In Eq. (5), Yi is described as a vector of length N_test, but it should be the vector of future values for sample i of length tau_f; also, Eq. (2) writes a squared norm with the notation ||.||_2. Please fix the notation throughout.
  3. [Section 4.4, Figure 7 text] The phrase '600 consecutive time periods randomly selected from the Inner Mongolia test set' is self-contradictory; please clarify whether the window is consecutive or randomly selected.
  4. [Section 4.3 and Section 4.4] The description of GPT4TS in Section 4.3 emphasizes that it does not use pre-trained word embeddings, while Section 4.4 says GPT4TS is fine-tuning the LLM backbone; please clarify which GPT4TS variant was implemented and whether the implementation follows the published method.
  5. [Section 4.5, Table 5] The text states that accuracy improves as the number of GPT-2 layers increases, but the table is not monotonic, for example Inner Mongolia Step 4 MAE is 6.41 for layer 2 versus 6.59 for layer 8; please rephrase the conclusion to say that performance is not significantly affected beyond a few layers.
  6. [General] No data or code availability statement is provided; given the evaluation-protocol concerns above, releasing the split indices and code would substantially improve reproducibility.
  7. [Section 2.1] The problem definition uses C stations for the historical input and M stations for the output without stating that M = C; in the experiments each region contains five wind farms, so please reconcile the notation.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the central claim is an empirical benchmark comparison, not a derivation that reduces to its inputs.

full rationale

The paper's central claim is that M2WLLM 'consistently outperforms existing methods, such as GPT4TS, across various datasets and prediction horizons' (Abstract, Section 4.4), supported by Table 2. This is an empirical benchmark result, not a formal derivation, so there is no equation-level chain in which a 'prediction' or 'first-principles result' is equivalent to its inputs by construction. The proposed architecture (Prompt Embedder, Data Embedder, Semantic Augmenter, LoRA) is described operationally; no fitted parameter is renamed as a prediction, no uniqueness theorem is imported, and no self-citation is invoked to justify the load-bearing claim. Some prior works cited are from overlapping or same-group authors (e.g., [18] H. Fan et al. and [22] M. Li et al.), but these appear only in the related-work discussion or as baselines whose numerical results are re-computed in this paper, not as evidence that forces the reported superiority. The more substantive evaluation issue is that the 3:1:1 split in Section 4.1 is not stated to be chronological, and with 96-step input windows this could create train/test leakage if random; however, this is a correctness/validity concern about the experiment, not an instance of circularity under the patterns enumerated here. Since no specific reduction to inputs can be quoted, the appropriate finding is no circularity.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The model introduces trainable neural components but no physical entities. The main unstated supports are the transferability of GPT-2 embeddings to numerical time series and the integrity of the data split.

free parameters (3)
  • Patch length and stride (l_p, s) = not reported
    Segmentation hyperparameters that control how much context each token carries; the paper derives P from l_p and s but never gives their values, and accuracy is sensitive to them.
  • LoRA rank r = not reported
    Rank of the low-rank adaptation; chosen by hand and not specified in the experimental setup, despite being a central component of fine-tuning.
  • Number of semantic augmenter heads H and mapped vocabulary size N_vM = not reported
    Cross-attention dimensions are defined but their values are omitted, leaving the main novelty under-specified.
assumptions (3)
  • domain assumption Pre-trained GPT-2 word embeddings contain transferable semantic knowledge that, when used as keys and values in cross-attention, helps regress future wind power values.
    Section 3.3.2 assumes the semantic augmenter can convert numerical patches into semantically meaningful representations; the paper does not independently test this assumption against equal-capacity non-semantic embeddings.
  • domain assumption Instance normalization and matching denormalization preserve the information needed for forecasting and do not leak information across samples.
    Section 3.3.1 applies per-sample normalization without discussing whether normalization statistics are computed using only the input window or the entire sample; if computed over the full sample including future values, it could leak.
  • ad hoc to paper The reported 3:1:1 split yields a fair test set for overlapping time series windows.
    Section 4.1 does not state whether split is chronological; if random, overlapping 96-step windows appear in both train and test.

how reviews work

0 comments
Cite this review

Pith. "Pith review of M2WLLM: Multi-Modal Multi-Task Ultra-Short-term Wind Power Prediction Algorithm Based on Large Language Model." pith.science (2026). https://pith.science/paper/EWUFKPWJ

@misc{pith2026250600531,
  author       = {Pith},
  title        = {Pith review of: M2WLLM: Multi-Modal Multi-Task Ultra-Short-term Wind Power Prediction Algorithm Based on Large Language Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EWUFKPWJ}},
  note         = {Machine review of arXiv:2506.00531}
}
read the original abstract

The integration of wind energy into power grids necessitates accurate ultra-short-term wind power forecasting to ensure grid stability and optimize resource allocation. This study introduces M2WLLM, an innovative model that leverages the capabilities of Large Language Models (LLMs) for predicting wind power output at granular time intervals. M2WLLM overcomes the limitations of traditional and deep learning methods by seamlessly integrating textual information and temporal numerical data, significantly improving wind power forecasting accuracy through multi-modal data. Its architecture features a Prompt Embedder and a Data Embedder, enabling an effective fusion of textual prompts and numerical inputs within the LLMs framework. The Semantic Augmenter within the Data Embedder translates temporal data into a format that the LLMs can comprehend, enabling it to extract latent features and improve prediction accuracy. The empirical evaluations conducted on wind farm data from three Chinese provinces demonstrate that M2WLLM consistently outperforms existing methods, such as GPT4TS, across various datasets and prediction horizons. The results highlight LLMs' ability to enhance accuracy and robustness in ultra-short-term forecasting and showcase their strong few-shot learning capabilities.

Figures

Figures reproduced from arXiv: 2506.00531 by the authors.

Figure 1
Figure 1. Demonstration of a pre-trained LLM utilizing token embedders and backbone models for text [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Overall architecture of the proposed model integrating diverse prompts and data [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Prompt for wind forecast task The prompt for the wind forecast task can be divided into three parts, namely the task prompt, the prompt for historical wind power data and the prompt for numerical weather prediction. The task prompt primarily outlines the target variables to be predicted, along with the time intervals and the lengths of the input and output data. The prompt for the historical wind power data can desc… view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Attention-inspired Semantic Augmenter The core mechanism of the semantic augmenter is to apply cross-attention between the patch data Xp and the pre-trained word embedding matrix E0 ∈ RNv0×dLLM , where Nv0 represents the vocabulary size, and dLLM is the hidden dimensio…
Figure 5
Figure 5. Figure 5: These parameters A and B can be optimized using gradient descent based on the training samples, as their dimensions are significantly smaller than those of W0. The final parameters W are obtained by summing these with the frozen parameters W0, expressed as W = W0 + BA.…
Figure 6
Figure 6. Figure 6: Distribution of daily RMSE on different datasets [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]
Figure 7
Figure 7. Figure 7: Forecast result of different methods 16 [PITH_FULL_IMAGE:figures/full_fig_p016_7.png]
Figure 8
Figure 8. Figure 8: Model performance with different amounts of training data [PITH_FULL_IMAGE:figures/full_fig_p017_8.png]
Figure 9
Figure 9. Figure 9: Training and inference time with different LLMs layers [PITH_FULL_IMAGE:figures/full_fig_p019_9.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

41 extracted references · 36 canonical work pages

  1. [1]

    Y. Xue, Y. Chen, J. Zhao, X. Liu, K. Li, W. Qiu, Y. Gang, A review on short-term and ultra-short-term wind power prediction, Dianli Xitong Zidonghua/Automation of Electric Power Systems 39 (6) (2015) 1–12

    work page 2015

  2. [2]

    S. Wang, J. Shi, W. Yang, Q. Yin, High and low frequency wind power prediction based on transformer and bigru-attention, Energy 288 (2024) 129753

    work page 2024

  3. [3]

    T. Hong, P. Pinson, Y. Wang, R. Weron, D. Yang, H. Zareipour, Energy forecasting: A review and outlook, IEEE Open Access Journal of Power and Energy 7 (2020) 376–388

    work page 2020

  4. [4]

    Q. Han, F. Meng, T. Hu, F. Chu, Non-parametric hybrid models for wind speed forecasting, Energy Conversion and Management 148 (2017) 554–568

    work page 2017

  5. [5]

    J. W. Messner, P. Pinson, Online adaptive lasso estimation in vector autoregres- sive models for high dimensional wind power forecasting, International Journal of Forecasting 35 (4) (2019) 1485–1498. 20

    work page 2019

  6. [6]

    Xiong, X

    X. Xiong, X. Guo, P. Zeng, R. Zou, X. Wang, A short-term wind power forecast method via xgboost hyper-parameters optimization, Frontiers in energy research 10 (2022) 905155

    work page 2022

  7. [7]

    S. Hu, Y. Xiang, H. Zhang, S. Xie, J. Li, C. Gu, W. Sun, J. Liu, Hybrid forecast- ing method for wind power integrating spatial correlation and corrected numerical weather prediction, Applied Energy 293 (2021) 116951

    work page 2021

  8. [8]

    X. Peng, Z. Yang, Y. Li, B. Wang, J. Che, Short-term wind power prediction based on stacked denoised auto-encoder deep learning and multi-level transfer learning, Wind Energy 26 (10) (2023) 1066–1081

    work page 2023

Show all 41 references
  1. [9]

    Sareen, B

    K. Sareen, B. K. Panigrahi, T. Shikhola, R. Sharma, An imputation and decompo- sition algorithms based integrated approach with bidirectional lstm neural network for wind speed prediction, Energy 278 (2023) 127799

    2023

  2. [10]

    Memarzadeh, F

    G. Memarzadeh, F. Keynia, A new short-term wind speed forecasting method based on fine-tuned lstm neural network and optimal input sets, Energy Conversion and Management 213 (2020) 112824

    2020

  3. [11]

    C. Yu, Y. Li, L. Zhao, Q. Chen, Y. Xun, A novel time-frequency recurrent network and its advanced version for short-term wind speed predictions, Energy 262 (2023) 125556

    2023

  4. [12]

    S.-X. Lv, L. Wang, Multivariate wind speed forecasting based on multi-objective fea- ture selection approach and hybrid deep learning model, Energy 263 (2023) 126100

    2023

  5. [13]

    H. Wu, K. Meng, D. Fan, Z. Zhang, Q. Liu, Multistep short-term wind speed fore- casting using transformer, Energy 261 (2022) 125231

    2022

  6. [14]

    Zhang, J

    H. Zhang, J. Yan, Y. Liu, Y. Gao, S. Han, L. Li, Multi-source and temporal attention network for probabilistic wind power prediction, IEEE Transactions on Sustainable Energy 12 (4) (2021) 2205–2218

    2021

  7. [15]

    X. Liu, J. Zhou, Short-term wind power forecasting based on multivariate/multi- step lstm with temporal feature attention mechanism, Applied Soft Computing 150 (2024) 111050

    2024

  8. [16]

    Z. Ma, G. Mei, A hybrid attention-based deep learning approach for wind power prediction, Applied Energy 323 (2022) 119608

    2022

  9. [17]

    C. Peng, Y. Zhang, B. Zhang, D. Song, Y. Lyu, A. Tsoi, A novel ultra-short-term wind power prediction method based on xa mechanism, Applied Energy 351 (2023) 121905

    2023

  10. [18]

    H.Fan, X.Zhang, S.Mei, K.Chen, X.Chen, M2gsnet: Multi-modalmulti-taskgraph spatiotemporal network for ultra-short-term wind farm cluster power prediction, Applied Sciences 10 (21) (2020) 7915

    2020

  11. [19]

    Liang, Q

    X. Liang, Q. Gu, X. You, Wpformer: A spatial-temporal graph transformer with auto-correlation for wind power forecasting, IEEE Transactions on Sustainable En- ergy (2024). 21

    2024

  12. [20]

    L. Ye, Y. Li, M. Pei, Y. Zhao, Z. Li, P. Lu, A novel integrated method for short- term wind power forecasting based on fluctuation clustering and history matching, Applied Energy 327 (2022) 120131

    2022

  13. [21]

    Von Krannichfeldt, Y

    L. Von Krannichfeldt, Y. Wang, T. Zufferey, G. Hug, Online ensemble approach for probabilistic wind power forecasting, IEEE Transactions on Sustainable Energy 13 (2) (2021) 1221–1233

    2021

  14. [22]

    M. Li, M. Yang, Y. Yu, M. Shahidehpour, F. Wen, Adaptive weighted combina- tion approach for wind power forecast based on deep deterministic policy gradient method, IEEE Transactions on Power Systems (2023)

    2023

  15. [23]

    X. Liu, Y. Zhang, Z. Zhen, F. Xu, F. Wang, Z. Mi, Spatio-temporal graph neural network and pattern prediction based ultra-short-term power forecasting of wind farm cluster, IEEE Transactions on Industry Applications (2023)

    2023

  16. [24]

    H. Xu, Y. Zhang, Z. Zhen, F. Xu, F. Wang, Adaptive feature selection and gcn with optimal graph structure-based ultra-short-term wind farm cluster power forecasting method, IEEE Transactions on Industry Applications (2023)

    2023

  17. [25]

    Y. Zhao, H. Liao, S. Pan, Y. Zhao, Interpretable multi-graph convolution network integrating spatio-temporal attention and dynamic combination for wind power fore- casting, Expert Systems with Applications 255 (2024) 124766

    2024

  18. [26]

    M. Yang, C. Ju, Y. Huang, Y. Guo, M. Jia, Short-term power forecasting of wind farm cluster based on global information adaptive perceptual graph convolution net- work, IEEE Transactions on Sustainable Energy (2024)

    2024

  19. [27]

    M. Yang, D. Wang, W. Zhang, X. Yv, A centralized power prediction method for large-scale wind power clusters based on dynamic graph neural network, Energy 310 (2024) 133210

    2024

  20. [28]

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

    2023

  21. [29]

    A. F. Ansari, L. Stella, C. Turkmen, X. Zhang, P. Mercado, H. Shen, O. Shchur, S. S. Rangapuram, S. P. Arango, S. Kapoor, et al., Chronos: Learning the language of time series, arXiv preprint arXiv:2403.07815 (2024)

    2024 arXiv

  22. [30]

    Gruver, M

    N. Gruver, M. Finzi, S. Qiu, A. G. Wilson, Large language models are zero-shot time series forecasters, Advances in Neural Information Processing Systems 36 (2024)

    2024

  23. [31]

    M. Jin, S. Wang, L. Ma, Z. Chu, J. Y. Zhang, X. Shi, P.-Y. Chen, Y. Liang, Y.-F. Li, S. Pan, Q. Wen, Time-LLM: Time series forecasting by reprogramming large language models, in: The Twelfth International Conference on Learning Represen- tations, 2024

    2024

  24. [32]

    Z. Li, L. Xia, J. Tang, Y. Xu, L. Shi, L. Xia, D. Yin, C. Huang, Urbangpt: Spatio- temporal large language models, in: Proceedings of the 30th ACM SIGKDD Con- ference on Knowledge Discovery and Data Mining, 2024, pp. 5351–5362. 22

    2024

  25. [33]

    X. Yu, Z. Chen, Y. Ling, S. Dong, Z. Liu, Y. Lu, Temporal data meets llm– explainable financial time series forecasting, arXiv preprint arXiv:2306.11025 (2023)

    2023 arXiv

  26. [34]

    Z. Lai, T. Wu, X. Fei, Q. Ling, Bert4st:: Fine-tuning pre-trained large language model for wind power forecasting, Energy Conversion and Management 307 (2024) 118331

    2024

  27. [35]

    T. Wu, Q. Ling, Stellm: Spatio-temporal enhanced pre-trained large language model for wind speed forecasting, Applied Energy 375 (2024) 124034

    2024

  28. [36]

    E. J. Hu, Y. Shen, P. Wallis, Z. Allen-Zhu, Y. Li, S. Wang, L. Wang, W. Chen, Lora: Low-rank adaptation of large language models, arXiv preprint arXiv:2106.09685 (2021)

    2021 arXiv

  29. [37]

    Hochreiter, Long short-term memory, Neural Computation MIT-Press (1997)

    S. Hochreiter, Long short-term memory, Neural Computation MIT-Press (1997)

    1997

  30. [38]

    C. Lea, M. D. Flynn, R. Vidal, A. Reiter, G. D. Hager, Temporal convolutional net- works for action segmentation and detection, in: proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017, pp. 156–165

    2017

  31. [39]

    Vaswani, Attention is all you need, Advances in Neural Information Processing Systems (2017)

    A. Vaswani, Attention is all you need, Advances in Neural Information Processing Systems (2017)

    2017

  32. [40]

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

    2021

  33. [41]

    H. Wu, J. Xu, J. Wang, M. Long, Autoformer: Decomposition transformers with auto-correlation for long-term series forecasting, Advances in neural information pro- cessing systems 34 (2021) 22419–22430. 23

    2021

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.