Pith. sign in

REVIEW 4 major objections 5 minor 34 references

VFEM: Visual Feature Empowered Multivariate Time Series Forecasting with Cross-Modal Fusion

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Frozen vision model reads series plots and beats time-series baselines

desk verdict Plausible architecture with consistently good numbers, but the central claim about the frozen vision encoder is not isolated; still worth a serious referee. read the letter →

arxiv 2510.03244 v2 pith:2WHHFKQO submitted 2025-09-25 cs.LG cs.AIcs.CV

Yanlong Wang , Hang Yu , Jian Xu , Fei Ma , Hongkang Zhang , Tongtong Feng , Zijian Zhang , Shao-Lun Huang
show 2 more authors
Danny Dongning Sun Xiao-Ping Zhang
This is my paper · ORCID
classification cs.LGcs.AIcs.CV
keywords multivariatetimeseriesforecastingcross-modalfusionlargevisionmodelchannel-independentarchitecturevisualizationspatiotemporalattentionparameter-efficientfine-tuningSigLip2
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 introduces a model called VIFO in the body text (VFEM on the title page) that renders a multivariate time series as a single image, with each variable as a row and each time step as a column, then feeds that image through a frozen large vision model. The authors claim this lets the vision encoder extract cross-variable patterns such as lead-lag relationships, shared periodicities, and holiday effects that channel-independent time series models cannot see. Those visual features are fused with a trainable temporal branch through cross-modal attention, and only 7.45% of the model's parameters are trained. On seven benchmark datasets and horizons of 96, 192, 336, and 720 steps, VIFO reports the lowest MSE and MAE among the compared baselines, with errors degrading more slowly at long horizons. The intended contribution is a parameter-efficient way to bring spatial pattern recognition to multivariate forecasting.

What carries the argument

The central object is the variable-sized image rendering of the multivariate series: rows are variables, columns are time steps, and pixel color encodes the normalized value at that point. The model uses a frozen SigLip2-base-NaFlex encoder because it accepts arbitrary image heights and widths, letting one vision backbone handle datasets with very different numbers of variables and a fixed look-back of 512 steps. A projection layer maps the vision encoder's patch embeddings to the hidden dimension, while a parallel temporal branch applies reversible instance normalization and a spatial-temporal attention module to the raw series. A cross-modal attention layer fuses the two hidden representations before a mapping layer produces the forecast. The parameter economy is the distinctive part: roughly 375 million parameters are frozen, and only about 30 million, or 7.45%, are trainable, covering the projection, temporal network, and fusion network.

What would settle it

Permute the row order of variables before rendering the image and re-run the same training and evaluation; if forecast error is unchanged, the visual branch is not using cross-variable spatial layout. Alternatively, take a dataset with hundreds of variables and a 720-step horizon and check whether the reported long-horizon advantage persists at that extreme aspect ratio.

Watch

Extended reading notes

Core claim

The paper's central claim is that a pre-trained vision model, kept frozen, can serve as a feature extractor for multivariate time series if the series is rendered as an M-by-L pixel image. The authors assert that this visual representation exposes patterns that are visible to the human eye and to a vision transformer but invisible to channel-independent forecasting architectures: daily and weekly periodicity, morning and evening rush hours, weekend effects, small holidays, and stable lead-lag relations between variables. VIFO combines that frozen visual branch with a trainable time-series branch using spatial-temporal attention, fuses the two representations with cross-modal attention, and maps the result to a forecast. The paper reports that this configuration achieves the best MSE and MAE on ETTh1, ETTh2, ETTm1, ETTm2, Electricity, Weather, and Traffic across all tested horizons, and that the visual branch matters most for long-horizon forecasting.

Load-bearing premise

The load-bearing premise is that rendering the multivariate series as a rows-by-columns image preserves the cross-variable relationships that matter for forecasting, and that a vision model pre-trained on natural images can extract them from that rendering.

Editorial extensions

If this is right

  • If the claim holds, channel-independent architectures are not the only efficient way to handle varying dataset dimensions; a frozen vision backbone can supply cross-variable structure without per-dataset full fine-tuning.
  • Because the vision encoder is frozen, the approach suggests that time series forecasting can inherit the scale and transfer properties of large vision models at a fraction of the training budget required by channel-dependent models.
  • The visual branch's larger contribution at long horizons implies that rendering-based encoders may be especially useful when forecasts extend beyond one or two seasonal cycles.
  • The variable-size encoding scheme extends to datasets with hundreds of variables, such as Traffic, without changing the architecture, which removes a common obstacle for channel-dependent designs.
  • The small trainable fraction means the method could be adapted to new domains by training only the projection, temporal, and fusion modules, making it a practical route for low-resource forecasting settings.

Reading between the lines

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

  • A direct test of whether the vision branch reads cross-variable structure rather than global texture would be to permute the row order of variables before rendering; if forecasts do not change, the encoder is not using inter-variable spatial layout.
  • The same rendering could be paired with image-style augmentations such as row or column shifts, color jitter, or patch masking to regularize the temporal branch, which the paper does not explore.
  • For datasets with extreme aspect ratios, such as hundreds of variables against 512 time steps, the variable-size encoder's patch grid may stretch the visual signal; checking whether the reported long-horizon advantage persists at that extreme would test the generality of the mechanism.
  • A comparison against the same temporal branch with a randomly initialized vision encoder of equal parameter count would isolate whether the gains come from pre-training on natural images or simply from having a large frozen feature extractor.
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

4 major / 5 minor

Summary. The manuscript proposes VIFO (also called VFEM in the abstract), a dual-branch multivariate time series forecasting model that renders the M×L input as an image, feeds it to a frozen SigLip2 vision encoder, and fuses the visual hidden states with a spatiotemporal-attention time-series branch via a cross-modal fusion layer. On seven benchmarks with look-back 512 and forecast horizons 96–720, the reported MSE and MAE are lower than all compared baselines, while only 7.45% of the model parameters are trainable. The central claim is that rendering multivariate series as images makes complex cross-channel patterns visible to a pretrained large vision model, yielding gains that channel-independent baselines cannot achieve.

Significance. If the central claim is validated, the result would be interesting: a frozen general-purpose vision encoder acting as a cross-variable feature extractor, combined with parameter-efficient fine-tuning, would offer a new and practical direction for multivariate forecasting. The choice of SigLip2 with variable input size is a sensible match to the large range of variable counts across datasets, and the reported 7.45% trainable-parameter fraction is a genuine efficiency feature. However, the paper's evidence does not yet isolate the proposed mechanism: the visual branch is ablated only as a whole, the rendering procedure is unspecified, and the reported differences are often small and lack uncertainty estimates. The significance of the central claim therefore depends on additional controls and reproducibility details.

major comments (4)
  1. [Sections 2.1 and 2.4] The image-generation process is not specified. Section 2.1 defines X_vs as an M×L×c image generated by rendering, but the paper never states how scalar series values are mapped to RGB channels, whether per-variable normalization or RevIN is applied before rendering, what colormap or value-range rescaling is used, or how the encoder handles images whose M or L is smaller than the vision encoder's patch size. For example, ETTh1 has M=7 and L=512, while Traffic has M=862 and L=512, so the geometry of the rendered image varies substantially across datasets. Because the central claim is that a pretrained natural-image encoder extracts meaningful cross-channel patterns from these renderings, the missing rendering recipe blocks reproducibility and leaves open the possibility that the visual branch exploits color-map artifacts rather than the intended cross-variable structure.
  2. [Table 3] The ablation study removes the entire visual pathway, so it cannot isolate the contribution of pretraining. The comparison between 'w/ all' and 'w/o VS modal' only shows that some visual pathway helps, not that a frozen SigLip2 LVM specifically is responsible; a randomly initialized vision encoder or a much smaller untrained encoder could plausibly yield similar small deltas. The paper should include controls that replace the frozen SigLip2 encoder with (a) a randomly initialized encoder of the same architecture, (b) a smaller vision encoder, and (c) a version with the image input permuted or corrupted, so that the claimed mechanism is distinguished from a simple increase in model capacity. Without such controls, the advertised 'pre-trained LVM' explanation is only one of several possible explanations of the observed gains.
  3. [Section 3.1 and Tables 2–3] No uncertainty estimates are reported even though the setup states that each experiment was run three times with different random seeds. Many of the reported differences are small enough that seed noise could change the ranking: for example, in Table 3 the ETTh1 horizon-96 MSE difference between 'w/o VS modal' (0.355) and 'w/ all' (0.348) is 0.007, and in Table 2 several comparisons between VIFO and the second-best baseline are of similar magnitude. Reporting means, standard deviations, or confidence intervals over the three seeds is necessary to support the claims that VIFO is best on all seven datasets and that each ablated module contributes meaningfully.
  4. [Section 2.5 and Section 1] The temporal modality branch is not channel-independent: it uses a spatiotemporal attention structure over segments, as described in Section 2.5 and reference [27]. Therefore the comparison with channel-independent baselines (PatchTST, GPT4TS, Chronos) does not by itself demonstrate that the visual pathway is the source of cross-variable gains. The observed improvements could in principle come from the spatiotemporal temporal branch already modeling cross-variable dependencies. The paper should explicitly compare against a variant of the same temporal branch that is channel-dependent without the visual branch, or otherwise show that the visual branch contributes cross-channel information beyond what the temporal branch already captures.
minor comments (5)
  1. [Abstract/title] The model name is inconsistent: the abstract and the beginning of the full text use 'VFEM', while the title, the rest of the full text, and the tables use 'VIFO'. One name should be used throughout.
  2. [Table 2] The table header lists F∈{96,129,336,720}; this should be F∈{96,192,336,720}. The same typo appears in the caption.
  3. [Table 2] Several cells lack spacing between consecutive numbers, for example '0.2940.330' in the ETTh2 horizon-96 row and '0.2040.248' in the Weather horizon-192 row. The formatting should be corrected.
  4. [Section 3.2] The sentence about long-sequence performance refers to 'forecasting length of 360' while the experiments use horizons 336 and 720; this is presumably a typo for 336.
  5. [Table 1] The caption and text state that '7.45% of its parameters' are trainable; this should be phrased as '7.45% of the total model parameters' for clarity, since Table 1 separately lists the frozen and trainable fractions.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the visual-transfer claim is an empirical, externally anchored claim rather than a definitional or fitted result.

full rationale

The paper's central claim is that rendering multivariate time series as an M x L image and passing it through a frozen SigLip2 encoder captures cross-channel patterns that channel-independent models miss (Sections 1, 2.2-2.4). This is an empirical transfer claim: the image is a deterministic rendering of X_ts, but the forecasting target (future values) is not defined in terms of the image or the visual features, and no parameter is fitted to the test outcome and then reported as a prediction. The visual encoder is external (SigLip2), frozen, and not fitted to the forecasting benchmarks, so the visual pathway's contribution is not forced by construction. The temporal branch adopts the authors' prior PSFormer architecture via citation [27], but it is used only as a component; the ablations in Table 3 show that removing either branch degrades performance, and the central novelty (cross-modal fusion with a frozen LVM) is not a restatement of PSFormer. No uniqueness theorem, ansatz-smuggling citation, or renaming of a known result is load-bearing. The under-specification of the rendering (color mapping, scaling, aspect ratios) and the absence of a random-encoder control are experimental limitations, not circularity.

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

The paper introduces no new physical or conceptual entities; VIFO is an architecture, not a postulated mechanism. No free constants are fitted to the forecast errors, so the free-parameter list is empty.

assumptions (3)
  • domain assumption Rendering multivariate time series as an M by L image with one row per variate preserves the cross-variable dependencies that matter for forecasting.
    This is the foundational premise of the visual branch, asserted in Section 2.2 and used in Section 2.4; no experiment directly verifies that the rendering, rather than the temporal branch, encodes these dependencies.
  • domain assumption A frozen SigLip2 encoder pretrained on natural images transfers to time-series heatmap-like images without fine-tuning.
    The model freezes the LVM (Section 3.1) and relies on transfer; the paper does not test unfrozen or smaller variants to confirm this premise.
  • domain assumption The SigLip2-base-Naflex variable-size encoder handles extreme aspect ratios such as 862 by 512 without significant degradation.
    The backbone is chosen because it allows arbitrary height and width (Section 3.1), but no experiment measures performance versus aspect ratio.

how reviews work

0 comments
Cite this review

Pith. "Pith review of VFEM: Visual Feature Empowered Multivariate Time Series Forecasting with Cross-Modal Fusion." pith.science (2026). https://pith.science/paper/2WHHFKQO

@misc{pith2026251003244,
  author       = {Pith},
  title        = {Pith review of: VFEM: Visual Feature Empowered Multivariate Time Series Forecasting with Cross-Modal Fusion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2WHHFKQO}},
  note         = {Machine review of arXiv:2510.03244}
}
read the original abstract

Large time series foundation models often adopt channel-independent architectures to handle varying data dimensions, but this design ignores crucial cross-channel dependencies. Meanwhile, existing cross-modal methods predominantly rely on textual modalities, leaving the spatial pattern recognition capabilities of vision models underexplored for time series analysis. To address these limitations, we propose VFEM, a cross-modal forecasting model that leverages pre-trained large vision models (LVMs) to capture complex cross-variable patterns. VFEM transforms multivariate time series into visual representations, enabling LVMs to perceive spatial relationships that are not explicitly modeled by channel-independent models. Through a dual-branch architecture, visual and temporal features are independently extracted and then fused via cross-modal attention, allowing complementary information from both modalities to enhance forecasting. By freezing the LVM and training only 7.45% of the total parameters, VFEM achieves competitive performance on multiple benchmarks, offering a new perspective on multivariate time series forecasting.

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 29 canonical work pages

  1. [27]

    TEST: Text prototype aligned embedding to activate llm’s ability for time series,

    Chenxi Sun, Yaliang Li, Hongyan Li, and Shenda Hong, “TEST: Text prototype aligned embedding to activate llm’s ability for time series,” arXiv preprint arXiv:2308.08241, 2023

    arXiv 2023

  2. [1]

    VFEM: Visual Feature Empowered Multivariate Time Series Forecasting with Cross-Modal Fusion

    INTRODUCTION Time series forecasting has been widely applied in diverse settings such as weather, power systems, transportation, and finance [1–4]. These scenarios often involve a wide variety of temporal data, where different time series frequently ex- hibit intricate interrelationships. Early time series forecast- ing methods focus on statistical models...

    work page Pith review arXiv

  3. [2]

    Problem Formulation As shown in Figure 2, the input multivariate time series is denoted asX ts ∈R M×L with look-back windowL: (x1, x2,

    METHODS 2.1. Problem Formulation As shown in Figure 2, the input multivariate time series is denoted asX ts ∈R M×L with look-back windowL: (x1, x2, . . . , xL)andMvariables, wherex t represents the M-dimensional vector at time stept. Additionally, the visual inputX vs∈R M×L×c is generated by rendering theM×L time series data as anM×Lpixel image, where eac...

    work page

  4. [3]

    EXPERIMENTS 3.1. Setup Since the number of variates in multivariate time series varies across different datasets, and the variable dimension and temporal dimension length of the input time series are often highly imbalanced, we adopt the encoder of SigLip2-base- Naflex [28] as the backbone of the vision large model, which allows arbitrary adjustment of im...

    work page

  5. [4]

    Specifically, it explores the dis- covery of structured spatiotemporal patterns from spatiotem- poral visualization graphs

    CONCLUSIONS This work investigates the application of visual representation of multivariate time series and cross-modal fusion in the field of time series forecasting. Specifically, it explores the dis- covery of structured spatiotemporal patterns from spatiotem- poral visualization graphs. By leveraging the feature extrac- tion capability of large vision...

    work page

  6. [5]

    End-to-end data-driven weather prediction,

    Anna Allen, Stratis Markou, Will Tebbutt, James Requeima, Wes- sel P Bruinsma, Tom R Andersson, Michael Herzog, Nicholas D Lane, Matthew Chantry, J Scott Hosking, et al., “End-to-end data-driven weather prediction,”Nature, vol. 641, no. 8065, pp. 1172–1179, 2025

    work page 2025

  7. [6]

    Time-varying pat- tern causality inference in global stock markets,

    Tao Wu, Xiangyun Gao, Sufang An, and Siyao Liu, “Time-varying pat- tern causality inference in global stock markets,”International Review of Financial Analysis, vol. 77, pp. 101806, 2021

    work page 2021

  8. [7]

    itransformer: Inverted transformers are effective for time series forecasting,

    Yong Liu, Tengge Hu, Haoran Zhang, Haixu Wu, Shiyu Wang, Lintao Ma, and Mingsheng Long, “itransformer: Inverted transformers are effective for time series forecasting,”arXiv preprint arXiv:2310.06625, 2023

    arXiv 2023

Show all 34 references
  1. [8]

    FourierGNN: Rethinking multivariate time series forecasting from a pure graph perspective,

    Kun Yi, Qi Zhang, Wei Fan, Hui He, Liang Hu, Pengyang Wang, Ning An, Longbing Cao, and Zhendong Niu, “FourierGNN: Rethinking multivariate time series forecasting from a pure graph perspective,” in Thirty-seventh Conference on Neural Information Processing Systems, 2023

    2023

  2. [9]

    Aut- oformer: Decomposition transformers with auto-correlation for long- term series forecasting,

    Haixu Wu, Jiehui Xu, Jianmin Wang, and Mingsheng Long, “Aut- oformer: Decomposition transformers with auto-correlation for long- term series forecasting,”Advances in neural information processing systems, vol. 34, pp. 22419–22430, 2021

    2021

  3. [10]

    Timesnet: Temporal 2d-variation modeling for gen- eral time series analysis,

    Haixu Wu, Tengge Hu, Yong Liu, Hang Zhou, Jianmin Wang, and Mingsheng Long, “Timesnet: Temporal 2d-variation modeling for gen- eral time series analysis,”arXiv preprint arXiv:2210.02186, 2022

    2022 arXiv

  4. [11]

    Fedformer: Frequency enhanced decomposed transformer for long-term series forecasting,

    Tian Zhou, Ziqing Ma, Qingsong Wen, Xue Wang, Liang Sun, and Rong Jin, “Fedformer: Frequency enhanced decomposed transformer for long-term series forecasting,” inInternational conference on ma- chine learning. PMLR, 2022, pp. 27268–27286

    2022

  5. [12]

    Crossformer: Transformer utilizing cross-dimension dependency for multivariate time series forecasting,

    Yunhao Zhang and Junchi Yan, “Crossformer: Transformer utilizing cross-dimension dependency for multivariate time series forecasting,” inThe Eleventh International Conference on Learning Representations, 2023

    2023

  6. [13]

    ModernTCN: A modern pure convolution structure for general time series analysis,

    Luo donghao and wang xue, “ModernTCN: A modern pure convolution structure for general time series analysis,” inThe Twelfth International Conference on Learning Representations, 2024

    2024

  7. [14]

    A time series is worth 64 words: Long-term forecast- ing with transformers,

    Yuqi Nie, Nam H. Nguyen, Phanwadee Sinthong, and Jayant Kalagnanam, “A time series is worth 64 words: Long-term forecast- ing with transformers,” 2023

    2023

  8. [15]

    MICN: Multi-scale local and global context modeling for long-term series forecasting,

    Huiqiang Wang, Jian Peng, Feihu Huang, Jince Wang, Junhui Chen, and Yifei Xiao, “MICN: Multi-scale local and global context modeling for long-term series forecasting,” inThe Eleventh International Con- ference on Learning Representations, 2023

    2023

  9. [16]

    Time-LLM: Time series forecasting by repro- gramming large language models,

    Ming Jin, Shiyu Wang, Lintao Ma, Zhixuan Chu, James Y . Zhang, Xi- aoming Shi, Pin-Yu Chen, Yuxuan Liang, Yuan-Fang Li, Shirui Pan, and Qingsong Wen, “Time-LLM: Time series forecasting by repro- gramming large language models,” inThe Twelfth International Con- ference on Learn...

  10. [17]

    One fits all: Power general time series analysis by pretrained lm,

    Tian Zhou, Peisong Niu, Liang Sun, Rong Jin, et al., “One fits all: Power general time series analysis by pretrained lm,”Advances in neu- ral information processing systems, vol. 36, pp. 43322–43355, 2023

    2023

  11. [18]

    DAM: Towards a foundation model for forecasting,

    Luke Nicholas Darlow, Qiwen Deng, Ahmed Hassan, Martin Asenov, Rajkarn Singh, Artjom Joosen, Adam Barker, and Amos Storkey, “DAM: Towards a foundation model for forecasting,” inThe Twelfth International Conference on Learning Representations

  12. [19]

    MOMENT: A family of open time-series foundation models,

    Mononito Goswami, Konrad Szafer, Arjun Choudhry, Yifu Cai, Shuo Li, and Artur Dubrawski, “MOMENT: A family of open time-series foundation models,” inForty-first International Conference on Ma- chine Learning, 2024

    2024

  13. [20]

    A decoder-only foundation model for time-series forecasting,

    Abhimanyu Das, Weihao Kong, Rajat Sen, and Yichen Zhou, “A decoder-only foundation model for time-series forecasting,”arXiv preprint arXiv:2310.10688, 2023

    2023 arXiv

  14. [21]

    UniTime: A language-empowered unified model for cross-domain time series forecasting,

    Xu Liu, Junfeng Hu, Yuan Li, Shizhe Diao, Yuxuan Liang, Bryan Hooi, and Roger Zimmermann, “UniTime: A language-empowered unified model for cross-domain time series forecasting,” inProceedings of the ACM on Web Conference 2024, 2024, pp. 4095–4106

    2024

  15. [22]

    Are language models actually useful for time series fore- casting?,

    Mingtian Tan, Mike Merrill, Vinayak Gupta, Tim Althoff, and Tom Hartvigsen, “Are language models actually useful for time series fore- casting?,”Advances in Neural Information Processing Systems, vol. 37, pp. 60162–60191, 2024

    2024

  16. [23]

    Unified training of universal time series forecasting transformers,

    Gerald Woo, Chenghao Liu, Akshat Kumar, Caiming Xiong, Silvio Savarese, and Doyen Sahoo, “Unified training of universal time series forecasting transformers,”arXiv preprint arXiv:2402.02592, 2024

    2024 arXiv

  17. [24]

    Timer: Generative pre-trained trans- formers are large time series models,

    Yong Liu, Haoran Zhang, Chenyu Li, Xiangdong Huang, Jianmin Wang, and Mingsheng Long, “Timer: Generative pre-trained trans- formers are large time series models,” 2024

    2024

  18. [25]

    TEMPO: Prompt-based generative pre-trained transformer for time series forecasting,

    Defu Cao, Furong Jia, Sercan O Arik, Tomas Pfister, Yixiang Zheng, Wen Ye, and Yan Liu, “TEMPO: Prompt-based generative pre-trained transformer for time series forecasting,” inThe Twelfth International Conference on Learning Representations, 2024

    2024

  19. [26]

    Time-LLM: Time series forecasting by repro- gramming large language models,

    Ming Jin, Shiyu Wang, Lintao Ma, Zhixuan Chu, James Y . Zhang, Xi- aoming Shi, Pin-Yu Chen, Yuxuan Liang, Yuan-Fang Li, Shirui Pan, and Qingsong Wen, “Time-LLM: Time series forecasting by repro- gramming large language models,” inThe Twelfth International Con- ference on Learn...

    2024

  20. [28]

    Timecma: Towards llm- empowered multivariate time series forecasting via cross-modality alignment,

    Chenxi Liu, Qianxiong Xu, Hao Miao, Sun Yang, Lingzheng Zhang, Cheng Long, Ziyue Li, and Rui Zhao, “Timecma: Towards llm- empowered multivariate time series forecasting via cross-modality alignment,” 2025

    2025

  21. [29]

    Moderntcn: A modern pure convolution structure for general time series analysis,

    Donghao Luo and Xue Wang, “Moderntcn: A modern pure convolution structure for general time series analysis,” inThe twelfth international conference on learning representations, 2024, pp. 1–43

    2024

  22. [30]

    Reversible instance normalization for accu- rate time-series forecasting against distribution shift,

    Taesung Kim, Jinhee Kim, Yunwon Tae, Cheonbok Park, Jang-Ho Choi, and Jaegul Choo, “Reversible instance normalization for accu- rate time-series forecasting against distribution shift,” inInternational Conference on Learning Representations, 2022

    2022

  23. [31]

    Psformer: Parameter-efficient transformer with segment attention for time series forecasting,

    Yanlong Wang, Jian Xu, Fei Ma, Shao-Lun Huang, Danny Dongning Sun, and Xiao-Ping Zhang, “Psformer: Parameter-efficient transformer with segment attention for time series forecasting,”arXiv preprint arXiv:2411.01419, 2025

    2025 arXiv

  24. [32]

    Siglip 2: Multilingual vision- language encoders with improved semantic understanding, localization, and dense features,

    Michael Tschannen, Alexey Gritsenko, Xiao Wang, Muhammad Ferjad Naeem, Ibrahim Alabdulmohsin, Nikhil Parthasarathy, Talfan Evans, Lucas Beyer, Ye Xia, Basil Mustafa, Olivier Hénaff, Jeremiah Harm- sen, Andreas Steiner, and Xiaohua Zhai, “Siglip 2: Multilingual vision- language...

    2025

  25. [33]

    Chronos: Learning the language of time series,

    Abdul Fatir Ansari, Lorenzo Stella, Ali Caner Turkmen, Xiyuan Zhang, Pedro Mercado, Huibin Shen, Oleksandr Shchur, Syama Sundar Ran- gapuram, Sebastian Pineda Arango, Shubham Kapoor, Jasper Zschieg- ner, Danielle C. Maddix, Hao Wang, Michael W. Mahoney, Kari Torkkola, Andrew G...

    2024

  26. [34]

    UniTST: Effectively modeling inter-series and intra-series dependencies for multivariate time series forecasting,

    Juncheng Liu, Chenghao Liu, Gerald Woo, Yiwei Wang, Bryan Hooi, Caiming Xiong, and Doyen Sahoo, “UniTST: Effectively modeling inter-series and intra-series dependencies for multivariate time series forecasting,”Transactions on Machine Learning Research, 2025

    2025

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.