Pith. sign in

REVIEW 4 major objections 6 minor 47 references

One Fits All: General Mobility Trajectory Modeling via Masked Conditional Diffusion

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read The paper introduces GenMove, a single masked conditional diffusion model that unifies trajectory generation, recovery, and prediction, and reports it outperforms task-specific baselines across six tasks.

desk verdict A plausible unified trajectory model whose central evaluation step is never described: the paper does not say how denoised embeddings become locations. read the letter →

arxiv 2501.13347 v1 pith:IF4FH6YA submitted 2025-01-23 cs.LG cs.AI

Qingyue Long , Can Rong , Huandong Wang , Yong Li This is my paper · ORCID
classification cs.LGcs.AI
keywords mobilitytrajectorymodelingconditionaldiffusionmodelsmaskedclassifier-freeguidancegenerationrecoverypredictionunifiedmodel
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

The paper tries to establish that one model can handle the main human-mobility trajectory tasks—generation, recovery, and prediction—instead of requiring a separate model for each. It argues that these tasks share common mobility patterns and that those patterns can be captured by a single masked conditional diffusion model, GenMove. The framework unifies task formats by masking the trajectory embedding, and it injects user context through a historical-embedding module with classifier-free guidance. If true, this would let urban planners, network operators, and mobility analysts train one model instead of several distinct systems. The paper reports GenMove outperforms state-of-the-art baselines across six tasks on two real-world datasets, with the largest improvement exceeding 13 percent in trajectory generation.

What carries the argument

The central object is the mask condition: a binary mask is applied element-wise to the trajectory embedding, splitting it into an observed portion that conditions the model and a target portion that the diffusion process must produce. The embedding itself comes from LINE, a network-embedding method applied to a spatial graph of visited locations. Five masking strategies—random, terminal, complete, sequential, and circadian rhythm—unify generation, recovery, and prediction into one format. A user embedding obtained from an LSTM over historical trajectories, or from a flow-based conditional controller, is injected into a transformer noise predictor through classifier-free guidance, letting the same network act conditional or unconditional. The reverse diffusion process follows the standard DDPM (denoising diffusion probabilistic model) update with classifier-free sampling. The mask is what lets one model switch between tasks.

What would settle it

Invert the pipeline: take a held-out trajectory, compute its embedding, add the same noise schedule, denoise it with the full trajectory as the mask condition, and decode; if the decoded location sequence does not match the original at a high rate, then the reported accuracy and distance metrics are measuring the decoder's behavior rather than the model's trajectory knowledge.

Watch

Extended reading notes

Core claim

The central claim is that trajectory generation, recovery, and prediction can be expressed as one conditional denoising problem. GenMove embeds each trajectory with a network-embedding method on a spatial graph of visited locations, masks part of the embedding to define the task's target, and trains a transformer-based noise predictor to reverse a diffusion process conditioned on the unmasked portion and on a user-level embedding from historical trajectories. For controllable generation, a flow-based conditional controller maps constraints such as radius of gyration to that user embedding. The same model therefore denoises random noise into a complete trajectory for generation, fills masked interior points for recovery, and extends the observed prefix for prediction. The paper reports that this single model matches or exceeds specialized baselines on all six tasks, including long-term prediction and scarcity-constrained prediction.

Load-bearing premise

The model's evaluation depends on the assumption that the denoised trajectory embedding can be faithfully turned back into a sequence of real locations, a step the paper does not describe or validate.

Editorial extensions

If this is right

  • One trained GenMove model can switch among unconditional generation, controllable generation, recovery, next-location prediction, long-term prediction, and scarcity-constrained prediction by changing the mask and condition.
  • The reported generation improvements above 13 percent indicate that a task-general model can learn shared mobility patterns at least as well as specialized generative models.
  • Because task performance peaks at a mask ratio below 1, mixing tasks during training can strengthen individual tasks rather than merely sharing a model.
  • Zero-shot results on unseen users show large gains in scarcity-constrained and long-term prediction, suggesting the model learns transferable mobility regularities.

Reading between the lines

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

  • A direct extension of the mask-conditioning idea would apply GenMove to other sequential spatiotemporal settings, such as taxi demand or animal movement, where generation, imputation, and forecasting share latent dynamics.
  • The paper's mask-ratio result suggests a concrete design experiment: make the mixture over masking strategies trainable and test whether per-task performance improves beyond the fixed ratios used here.
  • The zero-shot user results imply a natural cross-city test: train on one city's users, evaluate on another city's users without fine-tuning, and compare with city-specific models.
  • If the embedding decoder is reliable, adding more task types and mask strategies may improve every existing task, making the framework a candidate base for a trajectory foundation model.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes GenMove, a single masked conditional diffusion framework for multiple mobility trajectory tasks: generation, recovery, and prediction, including extended variants such as controllable generation, long-term prediction, and scarcity-constrained prediction. The method unifies task formats by applying binary masks to a LINE-based trajectory embedding, uses a transformer noise predictor, and incorporates task context via a user embedding combined through classifier-free guidance. Experiments on two proprietary Chinese mobility datasets (ISP, MME) compare against task-specific and general baselines, reporting performance that is often best or competitive across six tasks, with the largest gains in trajectory generation (around 13% average JSD improvement) and smaller gains in prediction and recovery.

Significance. If the central claim holds, GenMove would be a meaningful step toward a general trajectory foundation model, demonstrating that one architecture can serve several trajectory tasks through shared mobility patterns. The idea of unifying tasks via masking and contextual embeddings is timely and aligns with trends in other sequence domains. The paper also includes a zero-shot evaluation and a limited 'mutual enhancement' analysis, which are valuable. However, the evaluation is currently not reproducible: the inverse mapping from denoised embeddings to location sequences is never specified, key hyperparameters (mask mixing ratios, guidance scale, null-token drop probability) are omitted, and point estimates without variance are reported despite five runs. The strongest experimental evidence is in generation; the prediction improvements over DSTPP are marginal (e.g., ISP Acc@1 0.4248 vs 0.4201) and need statistical support. The contributions and the experimental scope are solid enough that these issues are fixable within a revision.

major comments (4)
  1. [Section IV-A and Section IV-D] The paper never specifies how the denoised trajectory embedding e0_ta is decoded back into a sequence of discrete locations. Section IV-A defines the mask operation on the LINE embedding eall (Eq. 7), and Section IV-D states that iterative denoising 'ultimately result[s] in clean data, denoted as e0_ta', but no decoder, nearest-neighbor lookup, or other inversion procedure is described. Since all reported metrics (JSD over Distance/Radius/Duration/Daily-loc/Density/Trip, Recall/MAP/Distance, Accuracy@k) are computed on location sequences, the entire experimental evaluation depends on this unspecified inverse mapping. The authors must provide the exact decoding procedure and justify that it does not introduce artifacts; otherwise the numbers in Tables II-IV and Figures 9-13 cannot be interpreted as measuring trajectory quality.
  2. [Section IV-A and Section V-B] The mask distribution g(·) is defined as a mixture of five masking strategies with 'adjustable ratios', and the guidance scale ω and null-token drop probability λ are introduced in Section IV-D, but none of these values are given in the paper. Table V lists architecture, learning rate, diffusion steps, and batch size, yet omits the mask mixing ratios, ω, and λ. This makes the experiments unrepeatable and prevents others from reproducing the reported improvements. Provide these hyperparameter values and, if they vary per dataset or task, report them per setting.
  3. [Section V-A3 and Tables II-IV] The paper states that 'We repeated the experiments for each task five times' and reports only the averages, without standard deviations, confidence intervals, or significance tests. This is particularly problematic for the prediction results, where the margin over the best baseline is small (e.g., ISP Acc@1: GenMove 0.4248 vs DSTPP 0.4201; MME: 0.4865 vs 0.4838). Without variance information, the claims of 'significantly outperforming' and of improvements in the 6-20% range (Section V-C, V-D) are not substantiated. The authors should supplement the point estimates with error bars or statistical tests.
  4. [Section III-B] The closed-form noising process is stated as xt = √αt x0 + (1 − αt) ε, which is not the standard DDPM expression and appears dimensionally inconsistent (the coefficient of ε should be √(1 − ᾱt)). The sampling equation (Eq. 13) uses both αt and ¯αt, while Eq. (1) defines αt as a cumulative product, creating confusion. If the implementation follows the standard DDPM, the text should state this clearly and correct Eqs. (1) and (11); if not, the training loss and sampling update need to be consistent with the actual noise schedule.
minor comments (6)
  1. [Section V-A4d] The text says the task-general baselines accomplish 'trajectory generation, recovery, and recovery'; the last word should be 'prediction'.
  2. [Section IV-D] The interpretation of ω is incorrect: with ω=0, Eq. (12) reduces to the conditional model weight ϵθ(·|eco, pu), not 'just conditional user embedding'; with ω=1, the model uses the standard linear guidance combination, not an unconditional model. Clarify the role of ω in classifier-free guidance.
  3. [Section IV-B] The sentence 'We obtain the historical trajectory embedding hu using the trajectory embedding method' is vague; specify whether the same LINE-based embedding is used for historical trajectories and how the LSTM input is formed.
  4. [Section V-D] The zero-shot results in Figure 13 are claimed to exceed the best baseline by 18-20%, but no numeric tables are provided and the procedure for adapting baselines to zero-shot evaluation is not described. Report the exact numbers and the baseline adaptation details.
  5. [Section I] There is a typo: 'meet user needs preciselys' should be 'meet user needs precisely'.
  6. [References and Figure readability] Reference [22] appears garbled, and several figures (e.g., Figures 9, 10, 13) are low-resolution and difficult to read; high-resolution figures and a cleaned reference list are needed.

Circularity Check

1 steps flagged · score 6.0 of 10

Conditioning for prediction/recovery is built from a full-trajectory embedding that already contains the masked target.

  1. self definitional [Section IV-A, Mask Condition, Eq. (7) and terminal strategy description]
    "we build a spatial graph G = ( V, E) that includes a set of nodes V that contains all visited locations, and the weight of each edge represents the Euclidean distance between locations. We utilize the LINE method [36] to obtain the current trajectory embedding eall. ... We transform eall into conditional observation eco and task target e0 ta through the following operations: ( eco = eall ⊙ m, e0 ta = eall ⊙ (1 − m), (7) ... Terminal strategy: This strategy masks future parts of a trajectory, such as the last point or a few points, and is particularly suited for prediction tasks."

    The prediction/recovery conditioning is produced by masking eall, but eall is defined as the LINE embedding of the full trajectory (the graph's node set contains 'all visited locations'). Therefore eco is not an embedding of the observed history; it is a masked slice of a representation fitted to the complete trajectory, including the future points that e0 ta is meant to predict. Target and condition are complementary parts of the same full-trajectory embedding. For the terminal mask used for prediction, the model is asked to recover the future part of an embedding that was already computed from that future, so the reported prediction/recovery gains are partly forced by construction rather than by generalization from genuinely observed locations.

full rationale

Aside from the loop above, I find no circular derivation. The mask ratios and classifier-free guidance weight are hyperparameters, not fitted to the test set; results are reported on held-out users. The heavy self-citations ([5], [9], [14], [20], [22], [34], [35], [43]) are prior design choices and baselines, not load-bearing uniqueness theorems; citing [34], [35] for diffusion-on-embedding does not by itself make the empirical claim circular. The paper's separate omission of any decoder from e0_ta to location sequences is a reproducibility gap, not a circular equivalence. Because the conditioning for prediction/recovery is constructed from the same full-trajectory embedding as the target, the central multi-task claim is partially circular; unconditional generation (complete mask) is not affected in the same way.

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

The central method relies on learned embeddings being invertible, on the standard classifier-free guidance training scheme, and on the discrete-time equal-interval trajectory definition. No new physical entities are introduced.

free parameters (3)
  • Mask mixing ratios
    The five masking strategies are sampled with adjustable ratios, but the ratios are not specified or justified. They control the balance of tasks during training.
  • Guidance scale omega
    The classifier-free guidance weight in Eq. 12 is a hyperparameter that trades sample quality for diversity, but its value is not reported.
  • Null token drop probability lambda
    The probability of setting user embedding to the null token for unconditional training is not reported or analyzed.
assumptions (3)
  • domain assumption Trajectory embeddings from LINE on the spatial graph encode enough information to reconstruct location sequences.
    The entire method runs diffusion on eall, but the inversion to locations is not explained.
  • domain assumption A single neural network can jointly parameterize conditional and unconditional diffusion with a null token.
    This is the standard classifier-free guidance training scheme, but it is an assumption about the model's capacity.
  • domain assumption Trajectories are recorded at equal time intervals and locations are discrete region IDs.
    Definition 1 in Section III-A states this, which limits applicability to continuous GPS or irregular sampling.

how reviews work

0 comments
Cite this review

Pith. "Pith review of One Fits All: General Mobility Trajectory Modeling via Masked Conditional Diffusion." pith.science (2026). https://pith.science/paper/IF4FH6YA

@misc{pith2026250113347,
  author       = {Pith},
  title        = {Pith review of: One Fits All: General Mobility Trajectory Modeling via Masked Conditional Diffusion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IF4FH6YA}},
  note         = {Machine review of arXiv:2501.13347}
}
read the original abstract

Trajectory data play a crucial role in many applications, ranging from network optimization to urban planning. Existing studies on trajectory data are task-specific, and their applicability is limited to the specific tasks on which they have been trained, such as generation, recovery, or prediction. However, the potential of a unified model has not yet been fully explored in trajectory modeling. Although various trajectory tasks differ in inputs, outputs, objectives, and conditions, they share common mobility patterns. Based on these common patterns, we can construct a general framework that enables a single model to address different tasks. However, building a trajectory task-general framework faces two critical challenges: 1) the diversity in the formats of different tasks and 2) the complexity of the conditions imposed on different tasks. In this work, we propose a general trajectory modeling framework via masked conditional diffusion (named GenMove). Specifically, we utilize mask conditions to unify diverse formats. To adapt to complex conditions associated with different tasks, we utilize historical trajectory data to obtain contextual trajectory embeddings, which include rich contexts such as spatiotemporal characteristics and user preferences. Integrating the contextual trajectory embedding into diffusion models through a classifier-free guidance approach allows the model to flexibly adjust its outputs based on different conditions. Extensive experiments on mainstream tasks demonstrate that our model significantly outperforms state-of-the-art baselines, with the highest performance improvement exceeding 13% in generation tasks.

Figures

Figures reproduced from arXiv: 2501.13347 by the authors.

Figure 1
Figure 1. Performance of GenMove on ISP dataset. approach involves using statistical interpolation or deep learning techniques to address missing data in trajectories [8], [9]. While these methods excel in specific tasks, they are limited by a lack of flexibility and generality. In fact, different trajectory tasks essentially rely on common human mobility patterns, such as periodicity, sequentiality, and circadian rhythms. Le… view at source ↗
Figure 2
Figure 2. The procedure of mobility modeling with GenMove. The reverse process [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The overview architecture of GenMove, which takes the noisy task target [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Different masking strategies are divided into task-specific masks and [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Conditional controller replaces History-based user embedding. [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 7
Figure 7. Figure 7: Six trajectory-related tasks, divided into basic and extended tasks. [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]
Figure 9
Figure 9. Figure 9: Comparison of heatmaps between real and generated trajectories on [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]
Figure 12
Figure 12. Figure 12: Performance of long-term trajectory prediction on two datasets. [PITH_FULL_IMAGE:figures/full_fig_p009_12.png]
Figure 11
Figure 11. Figure 11: Performance of scarcity-constrained trajectory prediction on two [PITH_FULL_IMAGE:figures/full_fig_p009_11.png]
Figure 14
Figure 14. Figure 14: Impact of the mask ratio on the performance of different tasks on [PITH_FULL_IMAGE:figures/full_fig_p010_14.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 32 canonical work pages

  1. [1]

    Generative ai empowered network digital twins: Architecture, technologies, and applications,

    T. Li, Q. Long, H. Chai, S. Zhang, F. Jiang, H. Liu, W. Huang, D. Jin, and Y . Li, “Generative ai empowered network digital twins: Architecture, technologies, and applications,” ACM Computing Surveys , 2025

    work page 2025

  2. [2]

    Public transport planning: When transit network connectivity meets commuting demand,

    S. Wang, Y . Sun, C. Musco, and Z. Bao, “Public transport planning: When transit network connectivity meets commuting demand,” in Proceedings of the 2021 International Conference on Management of Data , 2021, pp. 1906–1919

    work page 2021

  3. [3]

    Enhancing human mobility research with open and standardized datasets,

    T. Yabe, M. Luca, K. Tsubouchi, B. Lepri, M. C. Gonzalez, and E. Moro, “Enhancing human mobility research with open and standardized datasets,” Nature Computational Science , pp. 1–4, 2024

    work page 2024

  4. [4]

    Deepmove: Predicting human mobility with attentional recurrent networks,

    J. Feng, Y . Li, C. Zhang, F. Sun, F. Meng, A. Guo, and D. Jin, “Deepmove: Predicting human mobility with attentional recurrent networks,” in Proceedings of the 2018 world wide web conference , 2018, pp. 1459– 1468

    work page 2018

  5. [5]

    A Universal Model for Human Mobility Prediction

    Q. Long, Y . Yuan, and Y . Li, “A universal model for human mobility prediction,” arXiv preprint arXiv:2412.15294 , 2024

    work page Pith review arXiv 2024

  6. [6]

    The timegeo modeling framework for urban mobility without travel surveys,

    S. Jiang, Y . Yang, S. Gupta, D. Veneziano, S. Athavale, and M. C. González, “The timegeo modeling framework for urban mobility without travel surveys,” PNAS, vol. 113, no. 37, 2016

    work page 2016

  7. [7]

    Generating mobility trajectories with retained data utility,

    C. Cao and M. Li, “Generating mobility trajectories with retained data utility,” in Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining , 2021, pp. 2610–2620

    work page 2021

  8. [8]

    Network-less trajectory imputation,

    M. M. Elshrif, K. Isufaj, and M. F. Mokbel, “Network-less trajectory imputation,” in Proceedings of the 30th International Conference on Advances in Geographic Information Systems , 2022, pp. 1–10

    work page 2022

Show all 47 references
  1. [9]

    Attnmove: History enhanced trajectory recovery via attentional network,

    T. Xia, Y . Qi, J. Feng, F. Xu, F. Sun, D. Guo, and Y . Li, “Attnmove: History enhanced trajectory recovery via attentional network,” in Pro- ceedings of the AAAI Conference on Artificial Intelligence , vol. 35, no. 5, 2021, pp. 4494–4502

    2021

  2. [10]

    Representation learning and graph convolutional networks for short-term vehicle trajectory prediction,

    X. Xu, C. Yang, and W. Wu, “Representation learning and graph convolutional networks for short-term vehicle trajectory prediction,” Physica A: Statistical Mechanics and its Applications, vol. 637, p. 129560, 2024

    2024

  3. [11]

    Vehicle trajectory prediction and generation using lstm models and gans,

    L. Rossi, A. Ajmar, M. Paolanti, and R. Pierdicca, “Vehicle trajectory prediction and generation using lstm models and gans,” Plos one, vol. 16, no. 7, p. e0253868, 2021

    2021

  4. [12]

    Advancements in federated learning: Models, methods, and privacy,

    H. Chen, H. Wang, Q. Long, D. Jin, and Y . Li, “Advancements in federated learning: Models, methods, and privacy,” ACM Computing Surveys, vol. 57, no. 2, pp. 1–39, 2024

    2024

  5. [13]

    Diffusion model for gps trajectory generation,

    Y . Zhu, Y . Ye, X. Zhao, and J. J. Yu, “Diffusion model for gps trajectory generation,” arXiv preprint arXiv:2304.11582 , 2023

    2023 arXiv

  6. [14]

    Pategail: A privacy-preserving mobility trajectory generator with imitation learning,

    H. Wang, C. Gao, Y . Wu, D. Jin, L. Yao, and Y . Li, “Pategail: A privacy-preserving mobility trajectory generator with imitation learning,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 37, no. 12, 2023, pp. 14 539–14 547

    2023

  7. [15]

    Personalized route recommenda- tion using big trajectory data,

    J. Dai, B. Yang, C. Guo, and Z. Ding, “Personalized route recommenda- tion using big trajectory data,” in 2015 IEEE 31st international conference on data engineering . IEEE, 2015, pp. 543–554

    2015

  8. [16]

    A personalized recommendation framework with user trajectory analysis applied in location-based social network (lbsn),

    L. G. Xing, I. A. Abiodun, C. W. Khuen, and T. T. Boon, “A personalized recommendation framework with user trajectory analysis applied in location-based social network (lbsn),” in 2017 IEEE 3rd International Conference on Engineering Technologies and Social Sciences (ICETSS) ....

    2017

  9. [17]

    Vision-and-language pretrained models: A survey,

    S. Long, F. Cao, S. C. Han, and H. Yang, “Vision-and-language pretrained models: A survey,” arXiv preprint arXiv:2204.07356 , 2022

    2022 arXiv

  10. [18]

    Towards agi in computer vision: Lessons learned from gpt and large language models,

    L. Xie, L. Wei, X. Zhang, K. Bi, X. Gu, J. Chang, and Q. Tian, “Towards agi in computer vision: Lessons learned from gpt and large language models,” arXiv preprint arXiv:2306.08641 , 2023

    2023 arXiv

  11. [19]

    Gpt-4 technical report,

    J. Achiam, S. Adler, S. Agarwal, L. Ahmad, I. Akkaya, F. L. Aleman, D. Almeida, J. Altenschmidt, S. Altman, S. Anadkat et al. , “Gpt-4 technical report,” arXiv preprint arXiv:2303.08774 , 2023

    2023 arXiv

  12. [20]

    Learning to simulate human mobility,

    J. Feng, Z. Yang, F. Xu, H. Yu, M. Wang, and Y . Li, “Learning to simulate human mobility,” in Proc. KDD, 2020

    2020

  13. [21]

    Periodicmove: shift-aware human mobility recovery with graph neural network,

    H. Sun, C. Yang, L. Deng, F. Zhou, F. Huang, and K. Zheng, “Periodicmove: shift-aware human mobility recovery with graph neural network,” in Proceedings of the 30th ACM International Conference on Information & Knowledge Management , 2021, pp. 1734–1743

    2021

  14. [22]

    Privacy-preserving federated mobility prediction with com- pound data and model perturbation mechanism,

    L. Qingyue, W. Huandong, C. Huiming, J. Depeng, Z. Lin, Y . Li, and L. Yong, “Privacy-preserving federated mobility prediction with com- pound data and model perturbation mechanism,” China Communications, vol. 21, no. 3, pp. 160–173, 2024

    2024

  15. [23]

    Modelling the scaling properties of human mobility,

    C. Song, T. Koren, P. Wang, and A.-L. Barabási, “Modelling the scaling properties of human mobility,” Nature Physics, vol. 6, no. 10, pp. 818– 823, 2010. IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING, VOL. 36, NO. 12, JANUARY 2025 11

    2010

  16. [24]

    Mobtcast: Leveraging auxiliary trajectory forecasting for human mobility prediction,

    H. Xue, F. Salim, Y . Ren, and N. Oliver, “Mobtcast: Leveraging auxiliary trajectory forecasting for human mobility prediction,” Advances in Neural Information Processing Systems , vol. 34, pp. 30 380–30 391, 2021

    2021

  17. [25]

    Stan: Spatio-temporal attention network for next location recommendation,

    Y . Luo, Q. Liu, and Z. Liu, “Stan: Spatio-temporal attention network for next location recommendation,” in Proceedings of the web conference 2021, 2021, pp. 2177–2185

    2021

  18. [26]

    Sora: A review on background, technology, limitations, and opportunities of large vision models,

    Y . Liu, K. Zhang, Y . Li, Z. Yan, C. Gao, R. Chen, Z. Yuan, Y . Huang, H. Sun, J. Gao et al. , “Sora: A review on background, technology, limitations, and opportunities of large vision models,” arXiv preprint arXiv:2402.17177, 2024

    2024 arXiv

  19. [27]

    Timegpt-1,

    A. Garza and M. Mergenthaler-Canseco, “Timegpt-1,” arXiv preprint arXiv:2310.03589, 2023

    2023 arXiv

  20. [28]

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

    H. Wu, T. Hu, Y . Liu, H. Zhou, J. Wang, and M. Long, “Timesnet: Temporal 2d-variation modeling for general time series analysis,” arXiv preprint arXiv:2210.02186, 2022

    2022 arXiv

  21. [29]

    Simulating human mobility with a trajectory generation framework based on diffusion model,

    C. Chu, H. Zhang, P. Wang, and F. Lu, “Simulating human mobility with a trajectory generation framework based on diffusion model,” International Journal of Geographical Information Science , pp. 1–32, 2024

    2024

  22. [30]

    Time-llm: Time series forecasting by reprogramming large language models,

    M. Jin, S. Wang, L. Ma, Z. Chu, J. Y . Zhang, X. Shi, P.-Y . Chen, Y . Liang, Y .-F. Li, S. Panet al., “Time-llm: Time series forecasting by reprogramming large language models,” arXiv preprint arXiv:2310.01728, 2023

    2023 arXiv

  23. [31]

    Urbangpt: Spatio-temporal large language models,

    Z. Li, L. Xia, J. Tang, Y . Xu, L. Shi, L. Xia, D. Yin, and C. Huang, “Urbangpt: Spatio-temporal large language models,” arXiv preprint arXiv:2403.00813, 2024

    2024 arXiv

  24. [32]

    Csdi: Conditional score- based diffusion models for probabilistic time series imputation,

    Y . Tashiro, J. Song, Y . Song, and S. Ermon, “Csdi: Conditional score- based diffusion models for probabilistic time series imputation,” Advances in Neural Information Processing Systems , vol. 34, pp. 24 804–24 816, 2021

    2021

  25. [33]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” Advances in Neural Information Processing Systems , vol. 33, pp. 6840– 6851, 2020

    2020

  26. [34]

    Spatio-temporal diffusion point processes,

    Y . Yuan, J. Ding, C. Shao, D. Jin, and Y . Li, “Spatio-temporal diffusion point processes,” arXiv preprint arXiv:2305.12403 , 2023

    2023 arXiv

  27. [35]

    Towards generative modeling of urban flow through knowledge-enhanced denoising diffusion,

    Z. Zhou, J. Ding, Y . Liu, D. Jin, and Y . Li, “Towards generative modeling of urban flow through knowledge-enhanced denoising diffusion,” in Proceedings of the 31st ACM International Conference on Advances in Geographic Information Systems , 2023, pp. 1–12

    2023

  28. [36]

    Line: Large-scale information network embedding,

    J. Tang, M. Qu, M. Wang, M. Zhang, J. Yan, and Q. Mei, “Line: Large-scale information network embedding,” in Proceedings of the 24th international conference on world wide web , 2015, pp. 1067–1077

    2015

  29. [37]

    Summary of chatgpt-related research and perspective towards the future of large language models,

    Y . Liu, T. Han, S. Ma, J. Zhang, Y . Yang, J. Tian, H. He, A. Li, M. He, Z. Liu et al. , “Summary of chatgpt-related research and perspective towards the future of large language models,” Meta-Radiology, p. 100017, 2023

    2023

  30. [38]

    Human trajectory forecasting using a flow-based generative model,

    B. Zhang, T. Wang, C. Zhou, N. Conci, and H. Liu, “Human trajectory forecasting using a flow-based generative model,” Engineering Applica- tions of Artificial Intelligence , vol. 115, p. 105236, 2022

    2022

  31. [39]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” Advances in neural information processing systems , vol. 30, 2017

    2017

  32. [40]

    Dif- fwave: A versatile diffusion model for audio synthesis,

    Z. Kong, W. Ping, J. Huang, K. Zhao, and B. Catanzaro, “Dif- fwave: A versatile diffusion model for audio synthesis,” arXiv preprint arXiv:2009.09761, 2020

    2009 arXiv

  33. [41]

    Classifier-free diffusion guidance,

    J. Ho and T. Salimans, “Classifier-free diffusion guidance,” arXiv preprint arXiv:2207.12598, 2022

    2022 arXiv

  34. [42]

    Behavioral cloning from observation,

    F. Torabi, G. Warnell, and P. Stone, “Behavioral cloning from observation,” arXiv preprint arXiv:1805.01954 , 2018

    2018 arXiv

  35. [43]

    Practical synthetic human trajectories generation based on variational point processes,

    Q. Long, H. Wang, T. Li, L. Huang, K. Wang, Q. Wu, G. Li, Y . Liang, L. Yu, and Y . Li, “Practical synthetic human trajectories generation based on variational point processes,” in Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , 2023, pp....

    2023

  36. [44]

    Estimating human trajectories and hotspots through mobile phone data,

    S. Hoteit, S. Secci, S. Sobolevsky, C. Ratti, and G. Pujolle, “Estimating human trajectories and hotspots through mobile phone data,” Computer Networks, vol. 64, pp. 296–307, 2014

    2014

  37. [45]

    Reconstruction of human movement trajectories from large-scale low-frequency mobile phone data,

    M. Li, S. Gao, F. Lu, and H. Zhang, “Reconstruction of human movement trajectories from large-scale low-frequency mobile phone data,” Computers, Environment and Urban Systems , vol. 77, p. 101346, 2019

    2019

  38. [46]

    Next place prediction using mobility markov chains,

    S. Gambs, M.-O. Killijian, and M. N. del Prado Cortez, “Next place prediction using mobility markov chains,” in Proceedings of the first workshop on measurement, privacy, and mobility , 2012, pp. 1–6

    2012

  39. [47]

    Hst-lstm: A hierarchical spatial-temporal long- short term memory network for location prediction,

    D. Kong and F. Wu, “Hst-lstm: A hierarchical spatial-temporal long- short term memory network for location prediction,” in Twenty-Seventh International Joint Conference on Artificial Intelligence IJCAI-18 , 2018

    2018

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.