Pith. sign in

REVIEW 5 major objections 6 minor 1 cited by

PV-VLM: A Multimodal Vision-Language Approach Incorporating Sky Images for Intra-Hour Photovoltaic Power Forecasting

T0 review · 5 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read A multimodal model that fuses sky images, text prompts, and power history reports about 5–7% lower error in intra-hour solar forecasting.

desk verdict A plausible multimodal PV-forecasting architecture with modest reported gains, but the main evidence is compromised by an apparently random split and under-tuned baselines, so the actual improvement over a fair comparison is unproven. read the letter →

arxiv 2504.13624 v1 pith:KWSHIUSD submitted 2025-04-18 eess.SP

Huapeng Lin , Miao Yu This is my paper · ORCID
classification eess.SP
keywords photovoltaicpowerforecastingvision-languagemodelmultimodalfusionskyimagesintra-hourtransferlearningtimeserieslargelanguage
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 claims that intra-hour photovoltaic power forecasts can be made more accurate by combining three modalities: historical power readings, descriptive text prompts about the dataset and recent statistics, and sky images processed by a pretrained vision-language model. Existing methods typically process sky images and power series separately or with shallow feature concatenation, so they miss the deep interactions among cloud motion, system state, and domain knowledge. PV-VLM instead aligns all three modalities in a shared embedding space and lets the temporal representation query the vision-language representation through cross-modal attention. If the claim holds, grid operators and energy traders would gain a practical way to reduce 20-to-60-minute forecast errors, and the reported transfer results suggest the features keep working across climates without retraining. The headline numbers are roughly 5% RMSE and 6% MAE improvement on the primary dataset, and about 7% RMSE and 9.5% MAE improvement in a zero-shot transfer study.

What carries the argument

The load-bearing object is PV-VLM's three-branch encoder plus a cross-modal attention fusion. A PatchTST-style transformer divides the PV power history into overlapping patches and encodes both local fluctuations and long-range dependencies; a frozen pretrained vision-language encoder converts each sky image into semantic embeddings that highlight cloud edges, dense formations, and irradiance gradients; a language-model branch encodes a text prompt assembled from dataset descriptions and power statistics, together with a learnable soft prompt. The temporal embedding acts as the query in multi-head cross-modal attention, while the concatenated vision-language embedding acts as key and value, and a residual connection with layer normalization stabilizes the fused representation before a linear output head produces the forecast. This mechanism lets the time series selectively retrieve relevant visual and textual evidence for each forecast step rather than merely concatenating features.

What would settle it

Retrain all four baselines with the same hyperparameter search budget, early stopping, learning-rate schedule, and input lengths as PV-VLM on both datasets, and compare mean RMSE and MAE over the same three-run protocol; the claimed superiority fails if the tuned baselines become statistically indistinguishable or better. A sharper test is to replace the sky images with scrambled versions that preserve color statistics, or to remove the Vision-Aware Module while holding everything else fixed—if forecast error does not rise, the visual modality is not carrying the claimed information.

Watch

Extended reading notes

Core claim

PV-VLM claims to be the first systematic integration of a frozen pretrained vision-language encoder, a language-model prompt branch, and a PatchTST-style temporal encoder for intra-hour PV forecasting. On a 30-kW rooftop array dataset, it reports average RMSE reductions of about 5% and MAE improvements of nearly 6% over four deep-learning baselines across 20-, 40-, and 60-minute horizons. In a zero-shot transfer study to a second dataset in a different climate, it reports about 7% RMSE and 9.5% MAE reductions without any fine-tuning on the target site. The ablation study attributes the core of the gain to the temporal module, with text prompts and sky-image features adding complementary improvements, and a learnable soft prompt contributing smaller gains.

Load-bearing premise

The comparison treats the re-implemented baselines as fair competition even though their hyperparameters were not tuned for these datasets, while PV-VLM's hyperparameters were selected for these datasets; if the baselines were under-tuned or misconfigured, the reported error reductions would not show a genuine advantage.

Editorial extensions

If this is right

  • If the reported numbers hold, intra-hour solar forecasting at 20-to-60-minute horizons can be improved by roughly 5% RMSE and 6% MAE over current deep-learning baselines without any change to the physical PV plant.
  • Because the transfer study reports 7% RMSE and 9.5% MAE gains with no fine-tuning on the target site, the vision-language and prompt features appear to generalize across climates, PV sizes, and camera hardware.
  • The ablation shows the time-series history is the most load-bearing modality, with text prompts and sky images as complementary enhancements, so a partial deployment is possible when sky cameras are unavailable.
  • The learnable soft prompt preserves most of the full-model benefit, suggesting that prompt design contributes only a minor share of the overall improvement.
  • The framework introduces a concrete way to inject domain knowledge and dataset context into a forecasting model without changing its parameters: extend the text prompt with new statistics or descriptions.

Reading between the lines

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

  • Beyond the paper: the vision module's contribution is not isolated with a scrambled-image or random-image control, so it remains an open question how much of the gain comes from genuine cloud semantics rather than generic image statistics.
  • Beyond the paper: because the baselines were not hyperparameter-tuned for these datasets while the proposed model was, a fair equal-budget comparison might show a smaller real-world gap than the reported 5–9.5%.
  • Beyond the paper: the same three-branch architecture could be extended to wind, solar irradiance, or building-load nowcasting, where camera imagery and text metadata can be added as prompts.
  • Beyond the paper: the prompt branch offers a cheap route to inject exogenous information, such as weather alerts or market signals, into an already-trained model by simply extending the textual prompt.
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 / 6 minor

Summary. The paper introduces PV-VLM, a multimodal framework for intra-hour photovoltaic power forecasting that fuses time-series patches via a PatchTST-inspired transformer, textual prompts generated from historical statistics and dataset descriptors via an LLM, and sky-image features extracted by a pretrained vision-language model. The proposed architecture is evaluated on a Stanford dataset (Dataset A) and via a cross-dataset transfer study on two PV plants at the University of Wollongong (Dataset B). The authors report average RMSE reductions of approximately 5% on Dataset A and about 7% in the transfer study, with MAE improvements of nearly 6% and 9.5%, respectively, compared with four established baselines. The paper also includes a systematic ablation study examining the contribution of each modality and a visual heatmap illustration of the extracted sky-image features.

Significance. If the reported results are robust, PV-VLM makes a useful contribution to solar forecasting by demonstrating that a frozen vision-language model and an LLM-based prompt encoder can be combined with a time-series transformer for intra-hour predictions, and the cross-continental zero-shot transfer is a practically interesting result. The ablation study is systematic, and the heatmap visualization of SigLIP-2 feature activations is a nice qualitative check. However, the empirical claims are currently weakened by the ambiguous data split, the unequal hyperparameter tuning between the proposed model and baselines, and the absence of any uncertainty quantification or significance tests. These issues must be addressed before the reported relative improvements can be taken at face value.

major comments (5)
  1. [Section 3.2, data split] The paper states that a 70/10/20 allocation was used for training, validation, and testing, but it does not specify whether this split is chronological. Dataset A is a 2-minute resampled PV power time series with strong autocorrelation and cloud-event persistence over 20-60 minute horizons. Under a random split, test timestamps would have training samples from the same weather episodes both before and after them, allowing the model to interpolate or memorize rather than forecast the future. The authors should either confirm that the split is temporal (e.g., train on the first 70% of the time series, validate on the next 10%, and test on the final 20%) or provide the exact split indices so that temporal independence can be verified. If the split is random, the results in Tables 3 and 4, as well as the module attributions from the ablation study, may be substantially inflated.
  2. [Section 3.2, baseline tuning] The baselines were re-implemented 'strictly according to the methodological descriptions in their original publications, without additional dataset-specific adaptations or extensive hyperparameter tuning,' while the proposed PV-VLM's hyperparameters (Table 2) were evidently selected for these datasets. This asymmetry makes the comparison unfair: the reported 3.7-5.9% RMSE and 0.7-13.5% MAE improvements over the best baseline per horizon could partly reflect under-tuned baselines. The authors should perform a hyperparameter search for each baseline using the same validation set and computational budget, and report the best configuration for each method, or at least provide a sensitivity analysis showing that the conclusions are robust to baseline hyperparameters.
  3. [Section 3.2 and Tables 3-6, statistical significance] The paper reports that 'three experiments' were conducted and outcomes averaged, but no standard deviations or significance tests are provided anywhere. Given that some differences are very small (e.g., 0.71% MAE improvement over the best baseline at the 20-minute horizon in Table 3), the reported gains may not be statistically meaningful. The authors should report the mean and standard deviation across runs for each model and horizon, and perform a paired significance test (such as the Diebold-Mariano test or a paired t-test) to establish that the proposed method is significantly better than each baseline.
  4. [Section 3.1, data description] There is an inconsistency in the reported data sizes: the text states that Dataset A comprises approximately 349,000 samples after preprocessing, while Table 1 reports 62,910 samples; similarly, Dataset B is described as around 6,000 samples in the text but 3,000 in Table 1. Please clarify which numbers are correct and describe the preprocessing and quality-filtering steps, because the sample size directly affects the reliability of the comparisons and the transfer study results.
  5. [Section 3.5, transfer study setup] In the transfer study, the model is trained on Dataset A and tested on Dataset B without fine-tuning. The Prompt-Aware Module generates prompts from 'historical statistics and dataset descriptors.' If the dataset descriptor for the target dataset B (e.g., camera specifications or location) is included when constructing prompts at test time, the zero-shot evaluation is not purely agnostic to the target distribution. Please clarify whether the prompts are computed solely from the input time-series statistics (which is legitimate) or also from metadata of Dataset B, and if the latter, discuss the implications for the claimed zero-shot transfer.
minor comments (6)
  1. [Eq. (2-1)] The equation for bilinear interpolation is garbled in the manuscript (the summation notation and the description of coordinates are corrupted); please rewrite it in standard notation and define all variables.
  2. [Section 3.1] The statement that 'only the 2019 data were retained in Dataset A, matching Dataset B’s temporal coverage' is unclear because Dataset B's time range is not given; please specify the exact temporal windows of both datasets.
  3. [Fig. 4] The caption refers to '< > and < >' which appear to be missing placeholders; please complete the caption.
  4. [Tables 5 and 6] The units RMSE(10^4 kW) and RMSE(10^3 kW) are unusual and make cross-comparison difficult; please report all errors in kW consistently.
  5. [Section 3.4] In Table 4, the variant without the soft prompt gives a lower MAE than the full model at the 20-minute horizon (0.688 vs 0.698) and at the 40-minute horizon (0.845 vs 0.857), yet the text states that the soft prompt 'provides some minor improvements'; please reconcile this observation.
  6. [Reference [36]] Reference [36] is cited for Early Stopping, but that reference is a paper on sky-image-based forecasting, not a standard source for early stopping; please cite an appropriate reference or remove the citation.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper is an empirical, externally benchmarked comparison with no claim that reduces by construction to its inputs.

full rationale

PV-VLM is presented as a predictive model, not a derivation. The claimed 5%/6% Dataset A gains and 7%/9.5% transfer gains are measured against external baselines (MICNN-L, BILST, ConvLSTM, SUNSET) and an external Dataset B, so the central empirical claims do not reduce to the model's own fitted values. The Prompt-Aware Module builds prompts from historical statistics (range, median, trend) and dataset descriptors; these describe the input context, not the forecast target, and the validation is on held-out samples, so there is no fitted-input-called-prediction. The paper contains no self-citation chain or imported uniqueness theorem that forces its architecture; the citations to PatchTST, Time-LLM, CLIP/BLIP-2/SigLIP-2, and GPT-2 are standard external prior work, not load-bearing self-citations. Two experimental-validity concerns are present but are not circularity: the statement 'the baseline models were re-implemented strictly according to the methodological descriptions in their original publications, without additional dataset-specific adaptations or extensive hyperparameter tuning' raises fairness concerns, and 'A cross-validation strategy was applied, with 70% of the dataset allocated for training, 10% for validation, and 20% for testing' leaves open the possibility of a non-chronological split on autocorrelated 2-minute data. These are leakage/fairness risks, not definitional equivalence, and do not raise the circularity score.

Assumptions & free parameters 7 free parameters · 4 assumptions · 0 invented entities

The central claim rests on standard deep learning assumptions plus the domain-specific assumptions that sky images and textual statistics contain predictive signal for PV power. The most fragile items are the fairness of baseline re-implementations and the validity of the random split. No new physical entities or exotic parameters are introduced.

free parameters (7)
  • Initial learning rate eta_0 = 0.001
    Set in Table 2; not tuned per dataset in the paper.
  • Learning rate decay factor gamma = 0.1
    Set in Table 2.
  • Decay step size s = 3
    Set in Table 2.
  • Batch size = 32
    Set in Table 2.
  • Patience (early stopping) = 5
    Set in Table 2.
  • Patch length L_p and stride S
    Used in TAM patch segmentation (Eq. 2-9); exact values are not reported.
  • Model hidden dimension d_model and number of attention heads H
    Used in the fusion module (Eq. 2-12 to 2-16); dimensions are not specified.
assumptions (4)
  • domain assumption Pretrained VLM features from sky images are predictive of PV power fluctuations.
    VAM design in Section 2.1; only supported by a heatmap visualization, not by an isolated ablation.
  • domain assumption Textual prompts with historical statistics improve forecasting.
    PAM design in Section 2.2; ablation shows TAM+PAM is better than TAM alone, but the contribution is entangled with the soft prompt and fusion.
  • domain assumption The four baseline models can be fairly re-implemented from their publications.
    Section 3.2 states baselines were re-implemented without dataset-specific tuning; if this assumption fails, the comparison is invalid.
  • domain assumption The random 70/10/20 split produces independent test data.
    Section 3.2; for autocorrelated time series, random splits can leak nearby samples into training and inflate performance.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PV-VLM: A Multimodal Vision-Language Approach Incorporating Sky Images for Intra-Hour Photovoltaic Power Forecasting." pith.science (2026). https://pith.science/paper/KWSHIUSD

@misc{pith2026250413624,
  author       = {Pith},
  title        = {Pith review of: PV-VLM: A Multimodal Vision-Language Approach Incorporating Sky Images for Intra-Hour Photovoltaic Power Forecasting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KWSHIUSD}},
  note         = {Machine review of arXiv:2504.13624}
}
read the original abstract

The rapid proliferation of solar energy has significantly expedited the integration of photovoltaic (PV) systems into contemporary power grids. Considering that the cloud dynamics frequently induce rapid fluctuations in solar irradiance, accurate intra-hour forecasting is critical for ensuring grid stability and facilitating effective energy management. To leverage complementary temporal, textual, and visual information, this paper has proposed PV-VLM, a multimodal forecasting framework that integrates temporal, textual, and visual information by three modules. The Time-Aware Module employed a PatchTST-inspired Transformer to capture both local and global dependencies in PV power time series. Meanwhile, the Prompt-Aware Module encodes textual prompts from historical statistics and dataset descriptors via a large language model. Additionally, the Vision-Aware Module utilizes a pretrained vision-language model to extract high-level semantic features from sky images, emphasizing cloud motion and irradiance fluctuations. The proposed PV-VLM is evaluated using data from a 30-kW rooftop array at Stanford University and through a transfer study on PV systems at the University of Wollongong in Australia. Comparative experiments reveal an average RMSE reduction of approximately 5% and a MAE improvement of nearly 6%, while the transfer study shows average RMSE and MAE reductions of about 7% and 9.5%, respectively. Overall, PV-VLM leverages complementary modalities to provide a robust solution for grid scheduling and energy market participation, enhancing the stability and reliability of PV integration.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. CLIPTime: Time-Aware Multimodal Representation Learning from Images and Text

    cs.CV 2025-08 reject novelty 2.0 of 10

    CLIPTime adds a classification head and a transformer-style regression head to CLIP embeddings, hitting 98.7% accuracy on synthetic fungi but with weak timestamp predictions, especially for spores.

Reference graph

Works this paper leans on

36 extracted references · 15 canonical work pages · cited by 1 Pith paper

  1. [1]

    Solar energy: Potential and future prospects

    Kabir E, Kumar P, Kumar S, Adelodun AA, Kim K-H. Solar energy: Potential and future prospects. Renewable and Sustainable Energy Reviews 2018;82:894–900. https://doi.org/10.1016/j.rser.2017.09.094

    work page doi:10.1016/j.rser.2017.09.094 2018

  2. [2]

    The values of market -based demand response on improving power system reliability under extreme circumstances

    Wang F, Xu H, Xu T, Li K, Shafie -khah M, Catalão JoãoPS. The values of market -based demand response on improving power system reliability under extreme circumstances. Applied Energy 2017;193: 220–31. 21 https://doi.org/10.1016/j.apenergy.2017.01.103

    work page doi:10.1016/j.apenergy.2017.01.103 2017

  3. [3]

    Recurrent Spatial Pyramid CNN for Optical Flow Estimation

    Hu P, Wang G, Tan Y -P. Recurrent Spatial Pyramid CNN for Optical Flow Estimation. IEEE Trans Multimedia 2018;20:2814–23. https://doi.org/10.1109/TMM.2018.2815784

    work page arXiv 2018

  4. [4]

    Harvesting spatiotemporal correlation from sky image sequence to improve ultra -short-term solar irradiance forecasting

    Liu J, Zang H, Ding T, Cheng L, Wei Z, Sun G. Harvesting spatiotemporal correlation from sky image sequence to improve ultra -short-term solar irradiance forecasting. Renewable Energy 2023;209:619 –31. https://doi.org/10.1016/j.renene.2023.03.122

    work page doi:10.1016/j.renene.2023.03.122 2023

  5. [5]

    Pattern Classification and PSO Optimal Weights Based Sky Images Cloud Motion Speed Calculation Method for Solar PV Power Forecasting

    Zhen Z, Pang S, Wang F, Li K, Li Z, Ren H, et al. Pattern Classification and PSO Optimal Weights Based Sky Images Cloud Motion Speed Calculation Method for Solar PV Power Forecasting. IEEE Trans on Ind Applicat 2019;55:3331–42. https://doi.org/10.1109/TIA.2019.2904927

    arXiv 2019

  6. [6]

    Hybrid intrahour DNI forecast model based on DNI measurements and sky-imaging data

    Karout Y , Thil S, Eynard J, Guillot E, Grieu S. Hybrid intrahour DNI forecast model based on DNI measurements and sky-imaging data. Solar Energy 2023;249:541–58. https://doi.org/10.1016/j.solener.2022.11.032

    work page doi:10.1016/j.solener.2022.11.032 2023

  7. [7]

    Short -term solar radiatio n forecasting with a novel image processing -based deep learning approach

    Eşlik AH, Akarslan E, Hocaoğlu FO. Short -term solar radiatio n forecasting with a novel image processing -based deep learning approach. Renewable Energy 2022;200:1490–505. https://doi.org/10.1016/j.renene.2022.10.063

    work page doi:10.1016/j.renene.2022.10.063 2022

  8. [8]

    On vision transformer for ultra-short-term forecasting of photovoltaic generation using sky images

    Xu S, Zhang R, Ma H, Ekanayake C, Cui Y . On vision transformer for ultra-short-term forecasting of photovoltaic generation using sky images. Solar Energy 2024;267:112203. https://doi.org/10.1016/j.solener.2023.112203

    arXiv 2024

Show all 36 references
  1. [9]

    Intra -hour PV power forecasting based on sky imagery

    Paulescu M, Blaga R, Dughir C, Stefu N, Sabadus A, Calinoiu D, et al. Intra -hour PV power forecasting based on sky imagery. Energy 2023;279:128135. https://doi.org/10.1016/j.energy.2023.128135

    2023

  2. [10]

    Sky Image Prediction Model Based on Convolutional Auto-Encoder for Minutely Solar PV Power Forecasting

    Fu Y , Chai H, Zhen Z, Wang F, Xu X, Li K, et al. Sky Image Prediction Model Based on Convolutional Auto-Encoder for Minutely Solar PV Power Forecasting. IEEE Trans on Ind Applicat 2021;57 :3272–81. https://doi.org/10.1109/TIA.2021.3072025

    2021

  3. [11]

    Resampling and data augmentation for short-term PV output prediction based on an imbalanced sky images dataset using convolutional neural networks

    Nie Y , Zamzam AS, Brandt A. Resampling and data augmentation for short-term PV output prediction based on an imbalanced sky images dataset using convolutional neural networks. Solar Energy 2021;224:3 41–54. https://doi.org/10.1016/j.solener.2021.05.095

    2021 doi

  4. [12]

    CLIPSelf: Vision Transformer Distills Itself for Open -V ocabulary Dense Prediction

    Wu S, Zhang W, Xu L, Jin S, Li X, Liu W, et al. CLIPSelf: Vision Transformer Distills Itself for Open -V ocabulary Dense Prediction. The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024, OpenReview.net; 2024. https://doi....

    2024 doi arXiv

  5. [13]

    EdaDet: Open -V ocabulary Object Detection Using Early Dense Alignment

    Shi C, Yang S. EdaDet: Open -V ocabulary Object Detection Using Early Dense Alignment. Proceedings of the IEEE/CVF International Conference on C omputer Vision, 2023, p. 15724 –34. https://doi.org/10.48550/arXiv.2309.01151

    2023 doi arXiv

  6. [14]

    LLMs Meet VLMs: Boost Open V ocabulary Object Detection with Fine-grained Descriptors

    Jin S, Jiang X, Huang J, Lu L, Lu S. LLMs Meet VLMs: Boost Open V ocabulary Object Detection with Fine-grained Descriptors. The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024, OpenReview.net; 2024. https://doi.org/10.48...

    2024 doi arXiv

  7. [15]

    GPT4MTS: Prompt-based Large Language Model for Multimodal Time - series Forecasting

    Jia F, Wang K, Zheng Y , Cao D, Liu Y . GPT4MTS: Prompt-based Large Language Model for Multimodal Time - series Forecasting. Proce edings of the AAAI Conference on Artificial Intelligence 2024;38:23343 –51. https://doi.org/10.1609/aaai.v38i21.30383

    2024 doi

  8. [16]

    TEMPO: Prompt-based Generative Pre-trained Transformer for Time Series Forecasting

    Cao D, Jia F, Arik SÖ, Pfister T, Zheng Y , Ye W, et al. TEMPO: Prompt-based Generative Pre-trained Transformer for Time Series Forecasting. The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024, OpenReview.net; 2024. http...

    2024 doi arXiv

  9. [17]

    TEST: Text Prototype Aligned Embedding to Activate LLM’s Ability for Time Series

    Sun C, Li H, Li Y , Hong S. TEST: Text Prototype Aligned Embedding to Activate LLM’s Ability for Time Series. The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7 -11, 2024, OpenReview.net; 2023. https://doi.org/10.48550/arXiv.2308.08241

    2024 doi arXiv

  10. [18]

    A Time Series is Worth 64 Words: Long -term Forecasting with Transformers

    Nie Y , Nguyen NH, Sinthong P, Kalagnanam J. A Time Series is Worth 64 Words: Long -term Forecasting with Transformers. The Eleventh International Conference on Learning Representations, ICLR 2023, Kigali, Rwanda, May 1-5, 2023, OpenReview.net; 2023. https://doi.org/10.48550/a...

    2023 doi arXiv

  11. [19]

    Learning Transferable Visual Models From Natural Language Supervision

    Radford A, Kim JW, Hallacy C, Ramesh A, Goh G, Agarwal S, et al. Learning Transferable Visual Models From Natural Language Supervision. Proceedings of the 38th International Conference on Machine Learning, PMLR; 2021, p. 8748–63. https://doi.org/10.48550/arXiv.2103.00020

    2021 doi arXiv

  12. [20]

    BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models

    Li J, Li D, Savarese S, Hoi S. BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models. Proceedings of the 40th International Conference on Machine Learning, PMLR; 2023, p. 19730–42. https://doi.org/10.48550/arXiv.2301.12597

    2023 doi arXiv

  13. [21]

    SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features 202 5

    Tschannen M, Gritsenko A, Wang X, Naeem MF, Alabdulmohsin I, Parthasarathy N, et al. SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features 202 5. https://doi.org/10.48550/arXiv.2502.14786

    doi arXiv

  14. [22]

    Attention is all you need

    Vaswani A, Shazeer N, Parmar N, Uszkoreit J, Jones L, Gomez AN, et al. Attention is all you need. Proceedings of the 31st International Conference on Neural Information Processing Systems, Red Hook, NY , US A: Curran Associates Inc.; 2017, p. 6000–10. https://doi.org/10.48550/...

    2017 doi arXiv

  15. [23]

    Time-LLM: Time Series Forecasting by Reprogramming Large Language Models

    Jin M, Wang S, Ma L, Chu Z, Zhang JY , Shi X, et al. Time-LLM: Time Series Forecasting by Reprogramming Large Language Models. The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024, OpenReview.net; 2024

    2024

  16. [24]

    P-Tuning v2: Prompt Tuning Can Be Comparable to Fine -tuning Universally Across Scales and Tasks 2022

    Liu X, Ji K, Fu Y , Tam WL, Du Z, Yang Z, et al. P-Tuning v2: Prompt Tuning Can Be Comparable to Fine -tuning Universally Across Scales and Tasks 2022. https://doi.org/10.48550/arXiv.2110.07602

    2022 doi arXiv

  17. [25]

    Language Models are Unsupervised Multitask Learners

    Radford A, Wu J, Child R, Luan D, Amodei D, Sutskever I. Language Models are Unsupervised Multitask Learners. OpenAI 2019. https://cdn.openai.com/better -language- models/language_models_are_unsupervised_multitask_learners.pdf

    2019

  18. [26]

    BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

    Devlin J, Chang M-W, Lee K, Toutanova K. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. In: Burstein J, Doran C, Solorio T, editors. Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguis...

    2019 doi

  19. [27]

    Qwen2.5 Technical Report 2025

    Qwen, Yang A, Yang B, Zhang B, Hui B, Zheng B, et al. Qwen2.5 Technical Report 2025. https://doi.org/10.48550/arXiv.2412.15115

    2025 doi arXiv

  20. [28]

    Deep Residual Learning for Image Recognition

    He K, Zhang X, Ren S, Sun J. Deep Residual Learning for Image Recognition. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016, p. 770–8. https://doi.org/10.48550/arXiv.1512.03385

    2016 doi arXiv

  21. [29]

    SKIPP’D: A SKy Images and Photovoltaic Power Generation Dataset for short -term solar forecasting

    Nie Y , Li X, Scott A, Sun Y , Venugopal V , Brandt A. SKIPP’D: A SKy Images and Photovoltaic Power Generation Dataset for short -term solar forecasting. Solar Energy 2023;255:171 –9. https://doi.org/10.1016/j.solener.2023.03.043

    2023 doi

  22. [30]

    Sky Images and PV Power Measurements for Irradiance Forecasting 2021

    Dissawa L, Robinson D, Agalgaonkar A, Godaliyadda R, Ekanayake P, Perera S, et al. Sky Images and PV Power Measurements for Irradiance Forecasting 2021. https://doi.org/10.17632/cb8t8np9z3.2

    2021 doi

  23. [31]

    CAELUS: Classification of sky conditions from 1 -min time series of global solar irradiance using variability indices and dynamic thresholds

    Ruiz-Arias JA, Gueymard CA. CAELUS: Classification of sky conditions from 1 -min time series of global solar irradiance using variability indices and dynamic thresholds. Solar Energy 2023;263:111895. https://doi.org/10.1016/j.solener.2023.111895

    2023

  24. [32]

    Deep learning based solar radiation micro forecast by fusion of infrared cloud images and radiation data

    Ajith M, Martínez-Ramón M. Deep learning based solar radiation micro forecast by fusion of infrared cloud images and radiation data. Applied Energy 2021;294:117014. https://doi.org/10.1016/j.apenergy.2021.117014

    2021

  25. [33]

    Photovoltaic Nowcasting With Bi-Level Spatio-Temporal Analysis Incorporating Sky Images

    Zhang R, Ma H, Saha TK, Zhou X. Photovoltaic Nowcasting With Bi-Level Spatio-Temporal Analysis Incorporating Sky Images. IEEE Trans Sustain Energy 2021;12:1766–76. https://doi.org/10.1109/TSTE.2021.3064326

    2021

  26. [34]

    Sky image-based solar forecasting using deep learning with heterogeneous multi -location data: Dataset fusion versus transfer learning

    Nie Y , Paletta Q, Scott A, Pomares LM, Arbod G, Sgouridis S, et al. Sky image-based solar forecasting using deep learning with heterogeneous multi -location data: Dataset fusion versus transfer learning. Applied Energy 2024;369:123467. https://doi.org/10.1016/j.apenergy.2024....

    2024

  27. [35]

    Short-term solar power forecast with deep learning: Exploring optimal input and output configuration

    Sun Y , Venugopal V , Brandt AR. Short-term solar power forecast with deep learning: Exploring optimal input and output configuration. Solar Energy 2019;188:730–41. https://doi.org/10.1016/j.solener.2019.06.041

    2019 doi

  28. [36]

    Improving ultra-short-term photovoltaic power forecasting using a novel sky -image-based framework considering spatial -temporal feature interaction

    Zang H, Chen D, Liu J, Cheng L, Sun G, Wei Z. Improving ultra-short-term photovoltaic power forecasting using a novel sky -image-based framework considering spatial -temporal feature interaction. Energy 2024;293:130538. https://doi.org/10.1016/j.energy.2024.130538

    2024

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.