Pith. sign in

REVIEW 4 major objections 6 minor 3 cited by

T2S: High-resolution Time Series Generation with Text-to-Series Diffusion Models

T0 review · 4 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 single diffusion model can generate semantically aligned time series of arbitrary length from natural-language captions.

desk verdict T2S brings a useful new dataset and a clean variable-length latent design, but the current evaluation lets a model score well by memorizing its own captioning pipeline; the headline claims need a held-out caption test. read the letter →

arxiv 2505.02417 v2 pith:OZKFDEIP submitted 2025-05-05 cs.LG cs.AI

Yunfeng Ge , Jiawei Li , Yiji Zhao , Haomin Wen , Zhao Li , Meikang Qiu , Hongyan Li , Ming Jin
show 1 more author
Shirui Pan
This is my paper · ORCID
classification cs.LGcs.AI
keywords text-to-time-seriesgenerationdiffusionmodelsflowmatchingtransformerlength-adaptivevariationalautoencodertimeseriescaptionsTSFragment-600Karbitrary-length
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 argues that text-to-time-series generation can be moved out of single-domain settings. It introduces TSFragment-600K, a dataset of more than 600,000 fragment-level caption-series pairs, and T2S, a diffusion-based model that encodes variable-length series into a common latent space and conditions generation on natural-language captions using flow matching and a diffusion transformer. The authors report that T2S outperforms trained diffusion, variational-autoencoder, and zero-shot large-language-model baselines across 13 datasets from 12 domains, and that by training interleaved across lengths 24, 48, and 96 it can generate series of arbitrary length in that range from a single model. If these results hold, text prompts become a practical way to synthesize realistic temporal data in domains where data are scarce, imbalanced, or hard to collect.

What carries the argument

The load-bearing machinery is the combination of four pieces. The first is TSFragment-600K, a dataset built by cutting series into fragments, asking a language model to caption them with expert-curated seed prompts, and selecting among five candidate captions by embedding cosine similarity. The second is LA-VAE, a variational autoencoder that upsamples variable-length latent codes to a fixed size and downsamples them back, with a consistency loss on the latent code to reduce blur. The third is flow matching, a diffusion variant in which the denoiser predicts the straight-line velocity from noise to data along optimal-transport paths, integrated with a diffusion transformer (T2S-DiT) that injects caption information through adaptive layer normalization. The fourth is interleaved training, which shuffles samples of different lengths together so one model learns all lengths without catastrophic forgetting.

What would settle it

Run T2S on a held-out set of captions written by humans for time series the model has never seen, and compare its WAPE, MSE, and MRR@10 with the published TSFragment-600K numbers; a large drop would show the model is tied to its own captioning pipeline rather than to general natural-language understanding.

Watch

Extended reading notes

Core claim

The paper's central claim is that text and time series can be aligned at the fragment level by one domain-agnostic generative model. T2S first maps a variable-length series into a fixed-size latent embedding with a length-adaptive variational autoencoder (LA-VAE), then runs a diffusion transformer (T2S-DiT) trained with flow matching to denoise that embedding under a natural-language caption. The authors build TSFragment-600K, more than 600,000 fragment-level caption-series pairs drawn from eight classical time series datasets, and train T2S interleaved across lengths 24, 48, and 96 so that inference can produce any length in that range. They report that T2S obtains the best WAPE, MSE, and MRR@10 scores among trained diffusion, VAE, and zero-shot large-language-model baselines on 13 datasets spanning 12 domains, including point-level and instance-level captions as well as fragment-level ones.

Load-bearing premise

The benchmark assumes the evaluation captions test genuine language-to-series understanding, but the captions were produced from the same samples the model is later scored on, so the test may only measure how well the model inverts its own captioning pipeline.

Editorial extensions

If this is right

  • If T2S works as reported, a single model can generate series of any length within its trained range, so practitioners no longer need to train separate generators for each fixed sequence length.
  • The TSFragment-600K captioning pipeline gives other domains a recipe for building fragment-level text-series data, which is the missing ingredient for text-conditioned generation outside specialized fields.
  • Because T2S handles point-, fragment-, and instance-level captions, one model can serve tasks from fine-grained event descriptions to coarse global summaries.
  • The reported gains over diffusion- and LLM-based baselines suggest that the flow-matching and DiT design, not the caption size alone, drives generation quality across granularities.

Reading between the lines

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

  • The evaluation captions were generated from the same series used for scoring, so the strongest version of the claim is that T2S reproduces series from captions produced by its own dataset pipeline; a test with human-written captions on held-out series would separate language understanding from pipeline inversion.
  • The paper only demonstrates arbitrary-length generation inside the 24-96 range used in interleaved training; extrapolation to drastically longer horizons or to lengths not aligned with the trained multiples remains untested.
  • The model is evaluated on univariate series, so extending the same latent-space design to multivariate, multi-channel series would be a natural next test of the domain-agnostic claim.
  • A retrieval baseline that simply matches existing series to caption embeddings would clarify whether T2S truly generates new temporal structure or reproduces memorized fragments.
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 / 6 minor

Summary. The paper defines the text-to-time-series (T2S) generation task at three caption granularities (point, fragment, instance), introduces a new fragment-level dataset TSFragment-600K with over 600,000 GPT-generated caption-series pairs, and proposes a diffusion-based model T2S that combines a length-adaptive VAE (LA-VAE) with a Diffusion Transformer denoiser trained under flow matching and classifier-free guidance. Interleaved training across lengths is used to support generation at arbitrary lengths. The authors evaluate T2S on 13 datasets from 12 domains against DiffusionTS, TimeVAE, GPT-4o-mini, and Llama-3.1-8b, and claim state-of-the-art results on fragment-, point-, and instance-level benchmarks.

Significance. If the empirical claims are supported, the paper would make a useful contribution: it provides a large fragment-level text-time-series dataset, a domain-agnostic generation framework, and a length-adaptive latent VAE with a consistency loss. The theoretical framing is standard and internally consistent; there is no mathematical circularity in the flow-matching derivation. The release of code and dataset (GitHub link in Section 1) is a concrete strength. However, the current evaluation does not establish the central claim of semantically aligned generation from natural-language prompts, because the captions are generated from the same time series used for evaluation and the metrics measure fidelity to those original series. The 'arbitrary length' claim is also only tested at the three training lengths. These issues are fixable with additional experiments, but they are load-bearing for the stated SOTA claim.

major comments (4)
  1. [Section 2.2, Section 4.2, Eqs. (11)-(12)] The evaluation is circular with respect to the text-to-series semantic-alignment claim. TSFragment-600K captions are generated by GPT-4o-mini from the same time series fragments used for training and evaluation, and no train/test split or held-out human-authored caption set is described. WAPE and MSE in Eq. (11) directly compare generated series to the original y_i, and MRR@10 in Eq. (12) ranks candidates by cosine similarity to that same ground-truth y_i. A model that memorizes training series and uses the caption as a retrieval key can score well without demonstrating that it can satisfy a genuinely new caption. The authors should evaluate on held-out captions (e.g., human-written prompts or prompts not paired with any evaluated series) and report metrics that measure semantic match independently of the original series, such as retrieval against a caption-based latent space or human preference judgments.
  2. [Section 3.2, Table 1, Table 2] The 'arbitrary length' claim is only supported for the three training lengths {24, 48, 96}, plus a fixed length of 2048 for SUSHI. There is no evidence for interpolation to unseen lengths or extrapolation beyond the trained range. The paper should report generation quality at lengths not seen during training, e.g., 12, 72, 192, or explicitly restrict the claim to 'lengths within a specified range' with demonstrated coverage of that range.
  3. [Section 4.2, Table 1 footnote] The comparison to DiffusionTS and TimeVAE is potentially confounded by data budget. The footnote states that T2S uses interleaved training across lengths while baselines require fixed-length training and evaluation. If T2S sees all lengths of the same underlying series during training while each baseline sees only one length, T2S may benefit from more effective training data even without better text-series alignment. The authors should provide an ablation in which baselines receive the same interleaved or multi-length data, or otherwise control for the total amount of training data seen by each method.
  4. [Section 4.1, Tables 1-2] All reported numbers are point estimates with no error bars, multiple seeds, or significance tests. Given the small numeric gaps in some cells (e.g., MRR@10 values differing by 0.01-0.02 in Table 1), the SOTA claim is not statistically grounded. The authors should report mean and standard deviation over at least three independent runs and state whether the observed differences are significant.
minor comments (6)
  1. [Section 4.1, Eq. (12)] The definition of MRR@10 is incomplete: the threshold in cos(·) > threshold is never specified, and the procedure for forming the set of 10 generated candidates is not described. Please define the threshold and the candidate generation protocol.
  2. [Section 4.1, Point-Level Dataset] The description 'we adapted the dataset by concatenating each time series point with its associated text' is unclear. Please specify how point-level annotations are converted into training prompts and what sequence lengths result from this adaptation.
  3. [Section 3.3, Algorithm 1] The notation is inconsistent: line 12 iterates 'lengthi = 1 to k' but line 13 filters by len(s) = i, while earlier k denotes the number of datasets and lengths are denoted l1,...,ld. Please clarify whether the loop is over datasets or over lengths, and define how multiple lengths within one dataset are handled.
  4. [Section 2.2] The token limit ⟨M⟩ is introduced but its value is never given. Please report the actual token limit used for caption generation.
  5. [Section 3.2] The consistency-loss weight λ in Eq. (10) is not specified anywhere. Please report its value and the influence of λ on reconstruction quality.
  6. [Section 3.1] There are small typos: 'noise estimater' should be 'noise estimator' in the text following Eq. (4).

Circularity Check

1 steps flagged · score 6.0 of 10

Evaluation is partially circular: test captions are generated from the same time series used as ground truth, so WAPE/MSE/MRR@10 measure inversion of the captioning pipeline rather than generation from new captions.

  1. self definitional [Sec. 2.2 (TSFragment-600K construction) and Sec. 4.1 (Eq. 11, Eq. 12)]
    "First, a univariate time series x is segmented into k non-overlapping fragments, with each fragment x(j) ... for which textual captions are generated. ... These prompts guide GPT-4o-mini in generating concise, consistent, and semantically rich captions for all fragments. WAPE(y, ˆy) = P i∈Ω |yi− ˆyi| P i∈Ω |yi| ... ranki = argmin(n| cos(ˆyi,n,yi)> threshold) ... between the generated results and the truth."

    The caption for a test fragment is produced from that same fragment, so the reported metrics compare the model's output to the very series that generated the conditioning caption. WAPE and MSE are pointwise errors against y_i, and MRR@10 ranks candidates by cosine similarity to y_i. The benchmark therefore measures whether G(captioner(x)) approximates x, i.e. inversion of the caption-generation pipeline, not whether the model satisfies a genuinely new, never-paired caption. A model that memorizes training fragments and retrieves them by caption key can score well without semantic generalization.

full rationale

There is no circularity in the model derivation itself: LA-VAE, flow matching, DiT, and interleaved training are standard, independently motivated components. The circularity is confined to the empirical validation loop. TSFragment-600K constructs each caption from the very time series fragment that later serves as the ground-truth target, and every reported metric (WAPE, MSE, MRR@10) compares the generated series to that same target. Consequently, the reported numbers primarily measure how well the model inverts the GPT-4o-mini captioning process on the training distribution, rather than how well it generates series from genuinely new, human-authored captions. The comparison against DiffusionTS and TimeVAE is fair in the narrow sense that all methods face the same self-referential benchmark, but the absolute claim of semantic alignment and state-of-the-art capability is not independently supported. The arbitrary-length claim is also only tested at lengths 24, 48, and 96, which is an overgeneralization rather than a circularity. Overall, one central 'prediction' reduces by construction to an inversion task, warranting a score of 6.

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

The central claim rests on the reliability of automatically generated captions and the transferability of image-domain diffusion components; no new physical or conceptual entities are introduced.

free parameters (3)
  • lambda (consistency loss weight) = not reported
    Weight of the latent consistency loss in Eq. 10, chosen by the authors without a reported tuning procedure.
  • CFG scale and inference steps = CFG 7-10, steps 20-50
    Inference hyperparameters tuned on the Exchange Rate dataset (Figure 4) and applied across all datasets, which is a form of parameter fitting to a benchmark.
  • Training lengths = 24, 48, 96
    Lengths selected for interleaved training; the claim of arbitrary length generation is based only on this discrete set.
assumptions (3)
  • domain assumption GPT-4o-mini produces accurate and semantically aligned captions for time series fragments.
    Dataset construction in Section 2.2 depends on the captions being reliable, since they are used both for training and evaluation.
  • domain assumption Cosine similarity of text-embedding-3-small embeddings is a valid proxy for caption quality and semantic alignment.
    Section 2.2 uses average cosine similarity to select the best of five captions without human verification.
  • standard math Flow matching and DiT, standard in image generation, transfer to time series latent spaces.
    Section 3.1 adopts these methods from prior work without new theoretical justification.

how reviews work

0 comments
Cite this review

Pith. "Pith review of T2S: High-resolution Time Series Generation with Text-to-Series Diffusion Models." pith.science (2026). https://pith.science/paper/OZKFDEIP

@misc{pith2026250502417,
  author       = {Pith},
  title        = {Pith review of: T2S: High-resolution Time Series Generation with Text-to-Series Diffusion Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OZKFDEIP}},
  note         = {Machine review of arXiv:2505.02417}
}
read the original abstract

Text-to-Time Series generation holds significant potential to address challenges such as data sparsity, imbalance, and limited availability of multimodal time series datasets across domains. While diffusion models have achieved remarkable success in Text-to-X (e.g., vision and audio data) generation, their use in time series generation remains in its nascent stages. Existing approaches face two critical limitations: (1) the lack of systematic exploration of general-proposed time series captions, which are often domain-specific and struggle with generalization; and (2) the inability to generate time series of arbitrary lengths, limiting their applicability to real-world scenarios. In this work, we first categorize time series captions into three levels: point-level, fragment-level, and instance-level. Additionally, we introduce a new fragment-level dataset containing over 600,000 high-resolution time series-text pairs. Second, we propose Text-to-Series (T2S), a diffusion-based framework that bridges the gap between natural language and time series in a domain-agnostic manner. T2S employs a length-adaptive variational autoencoder to encode time series of varying lengths into consistent latent embeddings. On top of that, T2S effectively aligns textual representations with latent embeddings by utilizing Flow Matching and employing Diffusion Transformer as the denoiser. We train T2S in an interleaved paradigm across multiple lengths, allowing it to generate sequences of any desired length. Extensive evaluations demonstrate that T2S achieves state-of-the-art performance across 13 datasets spanning 12 domains.

Figures

Figures reproduced from arXiv: 2505.02417 by the authors.

Figure 1
Figure 1. Dataset Generation. GPT-4o-mini to generate high-quality [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overview of the T2S model. The framework conditions on captions for time series generation. LA-VAE encodes variable-length inputs into a latent space and decodes outputs to the original length. Forward diffusion transforms the original time series into noise, while T2S-DiT performs reverse denoising to align textual and temporal features, generating high-quality time series. The overall procedure can be formulated a… view at source ↗
Figure 4
Figure 4. Parameter sensitivity analysis on the Exchange Rate [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 3 Pith papers

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

  1. Overcoming the Modality Gap in Context-Aided Forecasting

    cs.LG 2026-03 unverdicted novelty 7.0 of 10

    A semi-synthetic dataset of 7 million context-augmented time series windows with verifier-filtered contexts enables transfer to real-world context-aided forecasting and suggests data quality—not architecture—was the b...

  2. Represent, Then Generate: Multimodal-Conditioned Time-Series Generation under Irregular Missingness

    cs.LG 2026-08 conditional novelty 6.0 of 10

    ReCoGen outperforms six baselines in downstream utility on all sixteen settings across three clinical datasets by decoupling condition representation (per-modality masked autoencoders) from generation (flow matching).

  3. Diff-MN: Diffusion Parameterized MoE-NCDE for Continuous Time Series Generation with Irregular Observations

    cs.LG 2026-01 conditional novelty 6.0 of 10

    Diff-MN generates continuous, arbitrary-resolution time series from irregular observations by diffusing MoE-NCDE dynamics weights, reporting consistent wins over KO-VAE and GT-GAN on ten datasets.

Reference graph

Works this paper leans on

42 extracted references · 15 canonical work pages · cited by 3 Pith papers

  1. [1]

    Diffusion-based conditional ecg gen- eration with structured state space models

    [Alcaraz and Strodthoff, 2023] Juan Miguel Lopez Alcaraz and Nils Strodthoff. Diffusion-based conditional ecg gen- eration with structured state space models. Computers in biology and medicine, 163:107115,

    work page 2023

  2. [4]

    Towards time series reasoning with llms

    [Chow et al., 2024] Winnie Chow, Lauren Gardiner, Haral- dur T Hallgr´ımsson, Maxwell A Xu, and Shirley You Ren. Towards time series reasoning with llms. arXiv preprint arXiv:2409.11376,

    arXiv 2024

  3. [5]

    Semeval-2017 task 5: Fine- grained sentiment analysis on financial microblogs and news

    [Cortis et al., 2017] Keith Cortis, Andr ´e Freitas, Tobias Daudert, Manuela Huerlimann, Manel Zarrouk, Siegfried Handschuh, and Brian Davis. Semeval-2017 task 5: Fine- grained sentiment analysis on financial microblogs and news. In Proceedings of the 11th international workshop on semantic evaluation (SemEval-2017) , pages 519–535,

    work page 2017

  4. [9]

    Scaling rectified flow transformers for high- resolution image synthesis

    [Esser et al., 2024] Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari, Jonas M ¨uller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boe- sel, et al. Scaling rectified flow transformers for high- resolution image synthesis. In Forty-first International Conference on Machine Learning,

    work page 2024

  5. [10]

    Creating synthetic energy meter data using conditional diffusion and building metadata

    [Fu et al., 2024] Chun Fu, Hussain Kazmi, Matias Quintana, and Clayton Miller. Creating synthetic energy meter data using conditional diffusion and building metadata. Energy and Buildings, 312:114216,

    work page 2024

  6. [11]

    Lumina-t2x: Transform- ing text into any modality, resolution, and duration via flow-based large diffusion transformers

    [Gao et al., 2024] Peng Gao, Le Zhuo, Ziyi Lin, Chris Liu, Junsong Chen, Ruoyi Du, Enze Xie, Xu Luo, Long- tian Qiu, Yuhang Zhang, et al. Lumina-t2x: Transform- ing text into any modality, resolution, and duration via flow-based large diffusion transformers. arXiv preprint arXiv:2405.05945,

    arXiv 2024

  7. [12]

    Emu video: Factorizing text-to-video gen- eration by explicit image conditioning

    [Girdhar et al., 2023] Rohit Girdhar, Mannat Singh, An- drew Brown, Quentin Duval, Samaneh Azadi, Sai Saketh Rambhatla, Akbar Shah, Xi Yin, Devi Parikh, and Is- han Misra. Emu video: Factorizing text-to-video gen- eration by explicit image conditioning. arXiv preprint arXiv:2311.10709,

    arXiv 2023

  8. [13]

    Denoising diffusion probabilistic models

    [Ho et al., 2020] Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. Advances in neural information processing systems , 33:6840–6851,

    2020

Show all 42 references
  1. [15]

    Putting the human in the time series analytics loop

    [Imani et al., 2019] Shima Imani, Sara Alaee, and Eamonn Keogh. Putting the human in the time series analytics loop. In Companion proceedings of the 2019 World Wide Web conference, pages 635–644,

    2019

  2. [17]

    Truth-conditional captioning of time series data

    [Jhamtani and Berg-Kirkpatrick, 2021] Harsh Jhamtani and Taylor Berg-Kirkpatrick. Truth-conditional captioning of time series data. arXiv preprint arXiv:2110.01839,

    2021 arXiv

  3. [18]

    Mimic-iv, a freely accessible electronic health record dataset

    [Johnson et al., 2023] Alistair EW Johnson, Lucas Bulgar- elli, Lu Shen, Alvin Gayles, Ayad Shammout, Steven Horng, Tom J Pollard, Sicheng Hao, Benjamin Moody, Brian Gow, et al. Mimic-iv, a freely accessible electronic health record dataset. Scientific data, 10(1):1,

    2023

  4. [19]

    Sushi: A system for unified semantic human in- teraction

    [Kawaguchi et al., 2025] Yohei Kawaguchi, Kota Dohi, and Aoi Ito. Sushi: A system for unified semantic human in- teraction. https://github.com/y-kawagu/SUSHI,

    2025

  5. [21]

    Diffusets: 12-lead ecg generation conditioned on clini- cal text reports and patient-specific information

    [Lai et al., 2025] Yongfan Lai, Jiabo Chen, Deyun Zhang, Yue Wang, Shijia Geng, Hongyan Li, and Shenda Hong. Diffusets: 12-lead ecg generation conditioned on clini- cal text reports and patient-specific information. arXiv preprint arXiv:2501.05932,

    2025 arXiv

  6. [22]

    V oice- box: Text-guided multilingual universal speech generation at scale

    [Le et al., 2024] Matthew Le, Apoorv Vyas, Bowen Shi, Brian Karrer, Leda Sari, Rashel Moritz, Mary Williamson, Vimal Manohar, Yossi Adi, Jay Mahadeokar, et al. V oice- box: Text-guided multilingual universal speech generation at scale. Advances in neural information processing...

    2024

  7. [23]

    Vector quantized time series generation with a bidi- rectional prior model

    [Lee et al., 2023] Daesoo Lee, Sara Malacarne, and Erlend Aune. Vector quantized time series generation with a bidi- rectional prior model. arXiv preprint arXiv:2303.04743 ,

    2023 arXiv

  8. [24]

    Flow matching for generative modeling

    [Lipman et al., 2022] Yaron Lipman, Ricky TQ Chen, Heli Ben-Hamu, Maximilian Nickel, and Matt Le. Flow matching for generative modeling. arXiv preprint arXiv:2210.02747,

    2022 arXiv

  9. [26]

    Time-mmd: A new multi-domain mul- timodal dataset for time series analysis

    [Liu et al., 2024a] Haoxin Liu, Shangqing Xu, Zhiyuan Zhao, Lingkai Kong, Harshavardhan Kamarthi, Aditya B Sasanur, Megha Sharma, Jiaming Cui, Qingsong Wen, Chao Zhang, et al. Time-mmd: A new multi-domain mul- timodal dataset for time series analysis. arXiv preprint arXiv:2406.08627,

    arXiv

  10. [27]

    Time weaver: A conditional time se- ries generation model

    [Narasimhan et al., 2024] Sai Shankar Narasimhan, Shub- hankar Agarwal, Oguzhan Akcin, Sujay Sanghavi, and Sandeep Chinchali. Time weaver: A conditional time se- ries generation model. arXiv preprint arXiv:2403.02682,

    2024

  11. [28]

    Gpt-4o mini,

    [OpenAI, 2023] OpenAI. Gpt-4o mini,

    2023

  12. [29]

    [Peebles and Xie, 2023] William Peebles and Saining Xie

    Accessed: 2025-01-23. [Peebles and Xie, 2023] William Peebles and Saining Xie. Scalable diffusion models with transformers. In Proceed- ings of the IEEE/CVF International Conference on Com- puter Vision, pages 4195–4205,

    2025

  13. [30]

    Movie gen: A cast of media foundation models

    [Polyak et al., 2024] Adam Polyak, Amit Zohar, Andrew Brown, Andros Tjandra, Animesh Sinha, Ann Lee, Apoorv Vyas, Bowen Shi, Chih-Yao Ma, Ching-Yao Chuang, et al. Movie gen: A cast of media foundation models. arXiv preprint arXiv:2410.13720,

    2024 arXiv

  14. [31]

    Astrom 3: A self-supervised multimodal model for astronomy

    [Rizhko and Bloom, 2024] Mariia Rizhko and Joshua S Bloom. Astrom 3: A self-supervised multimodal model for astronomy. arXiv preprint arXiv:2411.08842,

    2024 arXiv

  15. [32]

    High-resolution image synthesis with latent diffusion models

    [Rombach et al., 2022] Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn Ommer. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 10684–10695,

    2022

  16. [33]

    Exploring progress in mul- tivariate time series forecasting: Comprehensive bench- marking and heterogeneity analysis

    [Shao et al., 2024] Zezhi Shao, Fei Wang, Yongjun Xu, Wei Wei, Chengqing Yu, Zhao Zhang, Di Yao, Tao Sun, Guangyin Jin, Xin Cao, et al. Exploring progress in mul- tivariate time series forecasting: Comprehensive bench- marking and heterogeneity analysis. IEEE Transactions on K...

    2024

  17. [34]

    Medic: Mitigating eeg data scarcity via class-conditioned diffusion model

    [Sharma et al., 2023] Gulshan Sharma, Abhinav Dhall, and Ramanathan Subramanian. Medic: Mitigating eeg data scarcity via class-conditioned diffusion model. In Deep Generative Models for Health Workshop NeurIPS 2023 ,

    2023

  18. [35]

    Multi-modality conditional diffu- sion model for time series forecasting of live sales volume

    [Wang, 2024] Lijun Wang. Multi-modality conditional diffu- sion model for time series forecasting of live sales volume. In ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 2675–2679. IEEE,

    2024

  19. [36]

    Diffstg: Probabilistic spatio-temporal graph forecasting with denoising diffusion models

    [Wen et al., 2023] Haomin Wen, Youfang Lin, Yutong Xia, Huaiyu Wan, Qingsong Wen, Roger Zimmermann, and Yuxuan Liang. Diffstg: Probabilistic spatio-temporal graph forecasting with denoising diffusion models. InPro- ceedings of the 31st ACM International Conference on Ad- vance...

    2023

  20. [37]

    Context is key: A benchmark for forecasting with essen- tial textual information

    [Williams et al., 2024] Andrew Robert Williams, Arjun Ashok, ´Etienne Marcotte, Valentina Zantedeschi, Jithen- daraa Subramanian, Roland Riachi, James Requeima, Alexandre Lacoste, Irina Rish, Nicolas Chapados, et al. Context is key: A benchmark for forecasting with essen- tial...

    2024 arXiv

  21. [38]

    Autoformer: Decomposition transform- ers with auto-correlation for long-term series forecast- ing

    [Wu et al., 2021] Haixu Wu, Jiehui Xu, Jianmin Wang, and Mingsheng Long. Autoformer: Decomposition transform- ers with auto-correlation for long-term series forecast- ing. Advances in neural information processing systems , 34:22419–22430,

    2021

  22. [39]

    A survey on diffusion models for time series and spatio-temporal data

    [Yang et al., 2024] Yiyuan Yang, Ming Jin, Haomin Wen, Chaoli Zhang, Yuxuan Liang, Lintao Ma, Yi Wang, Chenghao Liu, Bin Yang, Zenglin Xu, et al. A survey on diffusion models for time series and spatio-temporal data. arXiv preprint arXiv:2404.18886,

    2024

  23. [40]

    Temporal data meets llm–explainable financial time series forecasting

    [Yu et al., 2023] Xinli Yu, Zheng Chen, Yuan Ling, Shujing Dong, Zongyi Liu, and Yanbin Lu. Temporal data meets llm–explainable financial time series forecasting. arXiv preprint arXiv:2306.11025,

    2023 arXiv

  24. [41]

    Diffusion-ts: Interpretable diffusion for general time series generation

    [Yuan and Qiao, 2024] Xinyu Yuan and Yan Qiao. Diffusion-ts: Interpretable diffusion for general time series generation. arXiv preprint arXiv:2403.01742,

    2024 arXiv

  25. [42]

    Unsupervised statistical feature-guided diffusion model for sensor-based human activity recognition

    [Zuo et al., 2023] Si Zuo, Vitor Fortes Rey, Sungho Suh, Stephan Sigg, and Paul Lukowicz. Unsupervised statistical feature-guided diffusion model for sensor-based human activity recognition. arXiv preprint arXiv:2306.05285 , 2023

    2023 arXiv

  26. [2009]

    Timevae: A variational auto-encoder for multivariate time series generation.arXiv preprint arXiv:2111.08095,

    [Desai et al., 2021] Abhyuday Desai, Cynthia Freeman, Zuhui Wang, and Ian Beaver. Timevae: A variational auto-encoder for multivariate time series generation.arXiv preprint arXiv:2111.08095,

    2021 arXiv

  27. [2017]

    Mean reciprocal rank

    [Craswell, 2009] Nick Craswell. Mean reciprocal rank. En- cyclopedia of database systems, pages 1703–1703,

    2009

  28. [2019]

    Clasp: Learning concepts for time-series signals from natural language supervision

    [Ito et al., 2024] Aoi Ito, Kota Dohi, and Yohei Kawaguchi. Clasp: Learning concepts for time-series signals from natural language supervision. arXiv preprint arXiv:2411.08397,

    2024 arXiv

  29. [2020]

    Arbitrary style transfer in real-time with adaptive instance normalization

    [Huang and Belongie, 2017] Xun Huang and Serge Be- longie. Arbitrary style transfer in real-time with adaptive instance normalization. In Proceedings of the IEEE in- ternational conference on computer vision , pages 1501– 1510,

    2017

  30. [2021]

    The llama 3 herd of models

    [Dubey et al., 2024] Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, et al. The llama 3 herd of models. arXiv preprint arXiv:2407.21783,

    2024 arXiv

  31. [2022]

    Flow straight and fast: Learning to generate and transfer data with rectified flow

    [Liu et al., 2022] Xingchao Liu, Chengyue Gong, and Qiang Liu. Flow straight and fast: Learning to generate and transfer data with rectified flow. arXiv preprint arXiv:2209.03003,

    2022 arXiv

  32. [2023]

    Timedit: General-purpose diffusion transform- ers for time series foundation model

    [Cao et al., 2024] Defu Cao, Wen Ye, Yizhou Zhang, and Yan Liu. Timedit: General-purpose diffusion transform- ers for time series foundation model. arXiv preprint arXiv:2409.02322,

    2024 arXiv

  33. [2024]

    Visionts: Visual masked autoencoders are free- lunch zero-shot time series forecasters

    [Chen et al., 2024] Mouxiang Chen, Lefei Shen, Zhuo Li, Xiaoyun Joy Wang, Jianling Sun, and Chenghao Liu. Visionts: Visual masked autoencoders are free- lunch zero-shot time series forecasters. arXiv preprint arXiv:2408.17253,

    2024 arXiv

  34. [2025]

    [Kong et al., 2020] Zhifeng Kong, Wei Ping, Jiaji Huang, Kexin Zhao, and Bryan Catanzaro

    Ac- cessed: 2025-01-02. [Kong et al., 2020] Zhifeng Kong, Wei Ping, Jiaji Huang, Kexin Zhao, and Bryan Catanzaro. Diffwave: A versa- tile diffusion model for audio synthesis. arXiv preprint arXiv:2009.09761,

    2025 arXiv

Pith tools

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