Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

MM-Path: Multi-modal, Multi-granularity Path Representation Learning -- Extended Version

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read MM-Path aims to show that combining road-network topology with remote sensing imagery yields generic path embeddings that beat single-modality and naive multi-modal baselines on travel-time estimation and path ranking.

desk verdict Novel idea and solid experimental scaffolding, but the two main loss equations as printed would train the opposite of the stated objective, and the unablated LightPath initialization muddies the gains. read the letter →

arxiv 2411.18428 v4 pith:KM6ZJXMG submitted 2024-11-27 cs.LG cs.AI

Ronghui Xu , Hanyin Cheng , Chenjuan Guo , Hongfan Gao , Jilin Hu , Sean Bin Yang , Bin Yang This is my paper · ORCID
classification cs.LGcs.AI
keywords pathrepresentationlearningmulti-modalremotesensingimageryroadnetworksmulti-granularityalignmentgraph-basedcross-modalfusiontraveltimeestimationranking
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 aims to establish that path representation learning improves when road-network topology is combined with remote sensing imagery, and that the right way to combine them is to align fine-to-coarse granularities before fusing through a graph. The proposed model, MM-Path, learns a single generic path embedding from two modalities: the road network and images of the region the path crosses. It reports consistent gains over all single-modal and multi-modal baselines on travel-time estimation and path ranking across two real-world city datasets, with improvements of roughly 5 to 13 percent. If correct, this gives transportation models a way to use visual context—such as road type and surrounding buildings—rather than relying on topology alone.

What carries the argument

The load-bearing pieces are the multi-granularity alignment loss, written as $\mathcal{L}_{\text{multi}} = \mathcal{L}_{\text{fine}} + \mathcal{L}_{\text{medium}} + \mathcal{L}_{\text{coarse}}$, and the graph-based cross-modal residual fusion loss $\mathcal{L}_{\text{fuse}}$. Fine alignment minimizes cosine distance between encoded node embeddings and the corresponding image patches; medium alignment does the same for road sub-path [sep] and image [sep] embeddings; coarse alignment is a bidirectional contrastive loss on path-level [cls] embeddings. Fusion concatenates each modality's initial embeddings with the other modality's encoded embeddings, builds a cross-modal adjacency matrix $\mathbf{A}$ whose modified identity excludes patches unrelated to any node, runs a two-layer GCN to update the residual embeddings, and applies a quadruplet loss that pulls the pooled road and image fused embeddings together relative to negatives.

What would settle it

Train MM-Path from random initialization and from the LightPath initialization with the same data, downstream heads, and hyperparameters; if the two versions perform within noise, the claimed contribution of the multi-modal components is not supported. A second check is to test whether the 5-13% gains persist when LightPath itself is retrained with the same extra compute and data.

Watch

Extended reading notes

Core claim

MM-Path is, to the authors' knowledge, the first path representation learning model to combine road-network data with remote sensing images into a single generic embedding. It encodes road paths and image paths in parallel Transformers, then aligns them at three granularities: node embeddings with image patches (fine), road sub-path embeddings with image embeddings (medium), and whole-path embeddings with whole image-sequence embeddings via contrastive loss (coarse). A graph-based cross-modal residual fusion component then builds a per-path directed graph whose edges encode intra-modal context, cross-modal spatial correspondence, and cross-modal context, runs a GCN over residual concatenations of initial and encoded embeddings from the two modalities, and uses a quadruplet loss to keep the fused road and image views consistent. The final representation is the concatenation of the two fused embeddings. Across travel-time estimation and path ranking on the Aalborg and Xi'an datasets, the model reports 5-13% improvements over the best baselines, with a case study showing image context resolving pairs of topologically similar paths.

Load-bearing premise

The paper loads the Road-Transformer from its own pre-trained LightPath model and never compares against training from scratch or against a version without that initialization, so the reported gains could come largely from the inherited representations rather than from the new alignment and fusion components.

Editorial extensions

If this is right

  • Path representation learning can move beyond topology-only models: adding remote sensing imagery yields measurable gains on both travel-time estimation and path ranking.
  • The fine-to-coarse alignment scheme gives a concrete way to handle granularity mismatch between sequence data and image data, not just for roads.
  • The graph-based residual fusion with a cross-modal adjacency matrix shows how spatial correspondence can guide fusion when modalities are heterogeneous.
  • Pre-trained MM-Path embeddings need less labeled data than training without pre-training, supporting use as a general-purpose path encoder.

Reading between the lines

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

  • If the LightPath-initialization confound is real, the marginal contribution of the multi-modal alignment and fusion may be smaller than the headline 5-13%; a from-scratch ablation would settle this.
  • The fixed image size and patch granularity were tuned on two cities; porting MM-Path to other cities likely requires re-selecting image resolution, and transfer tests could reveal whether the alignment losses generalize across map scales.
  • The same multi-granularity alignment plus graph fusion recipe could apply to street-view image sequences or other paired spatial-image data, where individual images map to segments of a sequence.
  • The authors' stated future focus on few-shot and zero-shot learning could be tested immediately by evaluating the generic embedding on a held-out city without fine-tuning.
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 / 5 minor

Summary. The paper proposes MM-Path, a pre-training framework that learns generic path representations by combining road-network topology with remote sensing imagery. The road path is encoded by a Transformer over node tokens, and the image path is encoded by a second Transformer over image patches. A multi-granularity alignment loss (Eqs. 5–7) aligns node/patch, sub-path/image, and path-level embeddings, while a graph-based cross-modal residual fusion component (Eqs. 8–12) fuses the two modalities through a GCN and a quadruplet loss. The final representation is the concatenation of the two fused embeddings. The authors report consistent improvements over single-modal and multi-modal baselines on travel time estimation and path ranking for Aalborg and Xi'an, with the best gains around 5–13%, and they release code.

Significance. If the empirical claims hold, MM-Path would be a useful step toward exploiting remote sensing imagery in path representation learning, and the combination of multi-granularity alignment with graph-based cross-modal fusion is a reasonable design direction. The paper is strengthened by experiments on two real-world datasets, a broad set of baselines, an ablation study, and a public code release. However, the two central loss functions as printed are inconsistent with the surrounding prose: Eq. 7 uses Euclidean distance as the similarity inside a contrastive exponential, and Eq. 12 minimizes the negative of a quadruplet hinge. Both would push matched pairs apart rather than together. Additionally, the Road-Transformer is initialized with the authors' own pre-trained LightPath, which is also the strongest single-modal baseline, and no from-scratch ablation is provided. These issues make the reported gains unattributable as written.

major comments (4)
  1. [§3.2.4, Eq. (7)] The coarse-grained contrastive loss is written with sim(pcls, hcls) defined as the Euclidean distance between pcls and hcls. Since exp(distance/σ) is large when the distance is large, minimizing the negative log-ratio in Eq. (7) rewards placing matched road-path and image-path [cls] embeddings far apart. This is the opposite of the stated alignment goal. The equation should use a quantity that is large for close vectors (for example, negative distance or cosine similarity), and the second fraction in the printed equation appears to omit the positive pair from its denominator; both need correction.
  2. [§3.3.3, Eq. (12)] The quadruplet loss is defined as the negative sum of two hinge terms. Because each term is nonnegative, minimizing Lfuse drives each hinge to be large, i.e., it forces ||y−z||² to exceed ||y−zN||² and ||z−yN||² by at least β. This is exactly the reverse of the intended constraint that the fused embeddings of the same path be closer than the fused embeddings of different paths. The leading minus sign should be removed (or the hinge terms should be subtracted inside a nonnegative loss), and the corrected form must be verified against the released implementation.
  3. [§4.1.2 and Table 2] The Road-Transformer is initialized with the authors' pre-trained LightPath, and LightPath is also the baseline against which the largest improvements are reported. No experiment trains MM-Path from scratch or otherwise ablates this initialization. Consequently, the 5–13% improvements in Tables 2–6 may reflect the warm-start representation inherited from LightPath rather than the proposed multi-granularity alignment or graph-based residual fusion. The authors should add a from-scratch (random-initialized or non-LightPath-initialized) MM-Path variant to the ablation study; without it, the marginal contribution of the new components is not identified.
  4. [Eq. (13) and Tables 2–6] Because Eqs. (7) and (12) are reversed as printed, the final objective in Eq. (13) cannot be the objective that produced the reported results unless the released code silently changes the signs. The paper should either correct the equations to match the implementation or explicitly document any difference between the printed objective and the actual training loss, and the corrected model should be re-evaluated on both datasets.
minor comments (5)
  1. [§4.2.3 and §4.2.4, Figures 5 and 6] The figures are difficult to interpret because the axis labels, legends, and curve labels appear as unicode path fragments (e.g., '/uni00000015/...') rather than readable text; the authors should replace these with plain-text labels.
  2. [§3.2.4, Eq. (7)] Using the name sim for a Euclidean distance is misleading; a distance should be negated or otherwise transformed if it is to act as a similarity in a contrastive objective.
  3. [§4.2.3, Figure 5] The claim that pre-training is beneficial would be stronger if the 'No Pre-trained' variant were also reported without the LightPath initialization, since the comparison conflates pre-training with the initialization choice.
  4. [§4.2.1, text after Table 2] There is a typo in the phrase 'TracjCL' in the discussion paragraph; it should read 'TrajCL' to match the baseline name in the table.
  5. [§3.2.1] The description of image position embeddings and road position embeddings is clear, but the relationship between the number of [sep] tokens and the number of images/sub-paths could be stated more explicitly to avoid ambiguity in the indexing of psep_i and hsep_i.

Circularity Check

1 steps flagged · score 4.0 of 10

The headline gain over LightPath is self-referential because MM-Path's Road-Transformer is initialized from the authors' own pre-trained LightPath and no from-scratch version is evaluated; the multi-modal alignment and fusion components otherwise have independent content.

  1. self citation load bearing [Section 4.1.2 (Implementation Details) and Table 2 (Overall Performance)]
    "To enhance the pre-training efficiency, we initialize our Road-Transformer with the pre-trained LightPath [45]."

    The proposed model's road encoder is initialized with the authors' own pre-trained LightPath, and the same LightPath is listed as a baseline in Table 2. No experiment removes or ablates this initialization, so the reported 'Improvement' over LightPath conflates the effect of the proposed multi-granularity alignment and graph-based fusion with simply continuing to train the LightPath representation. The comparison is self-referential: MM-Path contains LightPath by construction, so its superiority over LightPath is not an independent confirmation of the multi-modal contribution.

full rationale

The paper's formal derivation of the multi-granularity alignment losses (Eqs. 5–7) and graph-based cross-modal residual fusion (Eqs. 8–12) is self-contained: each loss is defined from the model's own embeddings and known spatial correspondences, and the final representation x = y||z is a concatenation of the two fused branch embeddings. The ablations in Tables 3–4 give some independent support for the alignment and fusion components. The main circularity-adjacent issue is the LightPath initialization, which is a self-citation of the authors' prior work and is load-bearing for the empirical comparison against LightPath; however, the central multi-modal idea still has independent content. Separately, the sign conventions in Eq. 12 (negative quadruplet hinge) and Eq. 7 (Euclidean distance in a contrastive exponent) appear to contradict the prose, but these are correctness risks, not circular reductions, and are not counted in the score beyond the self-citation concern.

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

The central empirical claim relies on standard deep learning components plus two debatable design choices: the LightPath initialization from the authors' prior work and the explicit removal of path-unrelated image patches. No new physical or conceptual entities are introduced; the cross-modal adjacency graph is an internal model construct, not an invented entity. The free parameters listed are typical hyperparameters, but their values are not always specified consistently.

free parameters (7)
  • margin beta in quadruplet loss (Eq. 12) = not specified
    Hyperparameter controlling the distance margin between positive and negative fused pairs; central to the fusion loss.
  • temperature sigma in coarse contrastive loss (Eq. 7) = learned
    Learned temperature scaling the softmax; value not reported.
  • loss weights lambda_mask, lambda_multi, lambda_fuse = 1, 1, 1
    Uniform weights chosen by hand in the training objective (Section 3.4).
  • mask ratio = 0.15
    Random masking ratio for the node masking task (Section 4.1.2).
  • patch grid per image = 4x4 = 16 patches
    Number of patches chosen for image path tokenization; sensitivity tested from 1x1 to 10x10 (Section 4.2.4).
  • embedding dimension d = 64 (MM-Path), 50 (baselines)
    The text sets d=64 for MM-Path and then says all methods are standardized to 50, an inconsistency that affects representation capacity comparisons.
  • number of Transformer layers = 5
    Depth of both Road-Transformer and Image-Transformer (Section 4.1.2).
assumptions (4)
  • domain assumption Remote sensing images and road network nodes are spatially aligned through coordinate mapping, so that each node lies in one image patch.
    Section 2.1 defines image paths by selecting images that correspond to latitudes and longitudes of nodes; the fine-grained alignment loss in Eq. 5 depends on this mapping being semantically meaningful.
  • ad hoc to paper Pre-trained LightPath weights provide a valid and fair initialization for the Road-Transformer branch.
    Section 4.1.2 states the Road-Transformer is initialized with pre-trained LightPath; MM-Path is then compared against LightPath as a baseline, so the initialization is an input from the authors' prior work rather than learned from scratch.
  • ad hoc to paper Patches that do not overlap any road node carry no useful path information and are excluded from the graph by zeroing their self-loop in the adjacency matrix.
    Section 3.3.2 introduces a modified identity matrix where diagonal entries for node-unrelated patches are set to 0; this design choice could discard relevant context such as buildings near the road.
  • standard math Standard Transformer and GCN architectures behave as described in the cited literature.
    The paper relies on Transformer blocks, GCN propagation (Eq. 8-9), and contrastive learning without re-deriving them.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MM-Path: Multi-modal, Multi-granularity Path Representation Learning -- Extended Version." pith.science (2026). https://pith.science/paper/KM6ZJXMG

@misc{pith2026241118428,
  author       = {Pith},
  title        = {Pith review of: MM-Path: Multi-modal, Multi-granularity Path Representation Learning -- Extended Version},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KM6ZJXMG}},
  note         = {Machine review of arXiv:2411.18428}
}
read the original abstract

Developing effective path representations has become increasingly essential across various fields within intelligent transportation. Although pre-trained path representation learning models have shown improved performance, they predominantly focus on the topological structures from single modality data, i.e., road networks, overlooking the geometric and contextual features associated with path-related images, e.g., remote sensing images. Similar to human understanding, integrating information from multiple modalities can provide a more comprehensive view, enhancing both representation accuracy and generalization. However, variations in information granularity impede the semantic alignment of road network-based paths (road paths) and image-based paths (image paths), while the heterogeneity of multi-modal data poses substantial challenges for effective fusion and utilization. In this paper, we propose a novel Multi-modal, Multi-granularity Path Representation Learning Framework (MM-Path), which can learn a generic path representation by integrating modalities from both road paths and image paths. To enhance the alignment of multi-modal data, we develop a multi-granularity alignment strategy that systematically associates nodes, road sub-paths, and road paths with their corresponding image patches, ensuring the synchronization of both detailed local information and broader global contexts. To address the heterogeneity of multi-modal data effectively, we introduce a graph-based cross-modal residual fusion component designed to comprehensively fuse information across different modalities and granularities. Finally, we conduct extensive experiments on two large-scale real-world datasets under two downstream tasks, validating the effectiveness of the proposed MM-Path. The code is available at: https://github.com/decisionintelligence/MM-Path.

Figures

Figures reproduced from arXiv: 2411.18428 by the authors.

Figure 1
Figure 1. A path in different modalities segments in paths, remote sensing images of paths (image paths for short) provide insights into geometric features and broader environ￾mental contexts (see [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. An example of image path processing Road Sub-paths. Given a road path R (𝑝) and an image path M (𝑝), the nodes of R (𝑝) located in the same image belong to a road sub￾path. Taken [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Overall framework of MM-Path embedding is obtained for each branch. We introduce a contrastive loss to ensure the consistency between the road fused embedding and image fused embedding. Finally, we concatenate these two fused embeddings to obtain a generic path representation. 3.2 Multi-granularity Alignment We model the road paths and image paths using Transformer ar￾chitecture, respectively. We then construct a mu… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: An example of multi-modal graph construction [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Effect of pre-training 1×1 2×2 4×4 5×5 10×10 Number of Patches 0.0 0.5 1.0 1.5 2.0 Testing (ms/path) Runtime 40 50 60 70 MAE MAE (a) Travel Time Estimation, Aalborg 1×1 2×2 4×4 5×5 10×10 Number of Patches 0.0 0.5 1.0 1.5 2.0 Testing (ms/path) Runtime 0.2 0.4 0.6 0.8 (b…
Figure 6
Figure 6. Figure 6: Effect of varying granularity size [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: Visualization of two paths 5 Related work 5.1 Path Representation Learning Models With the advancement of location-based services, the trajectory gen￾eration [51, 52] and trajectory analysis [6, 10, 13, 27, 28] have become increasingly prevalent. To better understand s…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Learning Generalized and Flexible Trajectory Models from Omni-Semantic Supervision

    cs.CV 2025-05 conditional novelty 6.0 of 10

    OmniTraj aligns four trajectory modalities (raw path, topology, road segments, regions) in a shared embedding space, enabling flexible condition-based retrieval that outperforms similarity-only baselines on Chengdu an...

Reference graph

Works this paper leans on

56 extracted references · 51 canonical work pages · cited by 1 Pith paper

  1. [1]

    Hangbo Bao, Li Dong, Songhao Piao, and Furu Wei. 2021. BEiT: BERT pre- training of image Transformers. In ICLR

    work page 2021

  2. [2]

    Hangbo Bao, Wenhui Wang, Li Dong, Qiang Liu, Owais Khan Mohammed, Kriti Aggarwal, Subhojit Som, Songhao Piao, and Furu Wei. 2022. Vlmo: Uni- fied vision-language pre-training with mixture-of-modality-experts. NeurIPS 35 (2022), 32897–32912

    work page 2022

  3. [3]

    Sean Bin Yang, Chenjuan Guo, Jilin Hu, Bin Yang, Jian Tang, and Christian S. Jensen. 2022. Weakly-supervised temporal path representation learning with contrastive curriculum learning. In ICDE. 2873–2885

    work page 2022

  4. [4]

    Shaosheng Cao, Wei Lu, and Qiongkai Xu. 2015. Grarep: Learning graph repre- sentations with global structural information. In CIKM. 891–900

    work page 2015

  5. [5]

    Yanchuan Chang, Jianzhong Qi, Yuxuan Liang, and Egemen Tanin. 2023. Con- trastive trajectory similarity learning with dual-feature attention. In ICDE. IEEE, 2933–2945

    work page 2023

  6. [6]

    Lu Chen, Yunjun Gao, Ziquan Fang, Xiaoye Miao, Christian S Jensen, and Chen- juan Guo. 2019. Real-time distributed co-movement pattern detection on streaming trajectories. PVLDB 12, 10 (2019), 1208–1220

    work page 2019

  7. [7]

    Lu Chen, Qilu Zhong, Xiaokui Xiao, Yunjun Gao, Pengfei Jin, and Christian S Jensen. 2018. Price-and-time-aware dynamic ridesharing. In ICDE. IEEE, 1061– 1072

    work page 2018

  8. [8]

    Meng Chen, Zechen Li, Weiming Huang, Yongshun Gong, and Yilong Yin. 2024. Profiling urban streets: A semi-supervised prediction model based on street view imagery and spatial topology. In KDD. 319–328

    work page 2024

Show all 56 references
  1. [9]

    Peng Chen, Yingying Zhang, Yunyao Cheng, Yang Shu, Yihang Wang, Qingsong Wen, Bin Yang, and Chenjuan Guo. 2024. Pathformer: Multi-scale transformers with adaptive pathways for time series forecasting. In ICLR

    2024

  2. [10]

    Yile Chen, Gao Cong, and Cuauhtemoc Anda. 2023. Teri: An effective framework for trajectory recovery with irregular time intervals. PVLDB 17, 3 (2023), 414– 426

    2023

  3. [11]

    Yile Chen, Xiucheng Li, Gao Cong, Zhifeng Bao, Cheng Long, Yiding Liu, Arun Kumar Chandran, and Richard Ellison. 2021. Robust road network rep- resentation learning: When traffic patterns meet traveling semantics. In CIKM. 211–220

    2021

  4. [12]

    Jian Dai, Bin Yang, Chenjuan Guo, and Zhiming Ding. 2015. Personalized route recommendation using big trajectory data. In ICDE. IEEE, 543–554

    2015

  5. [13]

    Xin Ding, Lu Chen, Yunjun Gao, Christian S Jensen, and Hujun Bao. 2018. UlTraMan: A unified platform for big trajectory data management and analytics. PVLDB 11, 7 (2018), 787–799

    2018

  6. [14]

    Ahmed Elbery, Hossam S Hassanein, Nizar Zorba, and Hesham A Rakha. 2020. Iot-based crowd management framework for departure control and navigation. IEEE Trans. Veh. Technol.70, 1 (2020), 95–106

    2020

  7. [15]

    Noel Gorelick, Matt Hancher, Mike Dixon, Simon Ilyushchenko, David Thau, and Rebecca Moore. 2017. Google Earth Engine: Planetary-scale geospatial analysis for everyone. Remote Sens. Environ. (2017)

    2017

  8. [16]

    Aditya Grover and Jure Leskovec. 2016. Node2vec: Salable feature learning for networks. In KDD. 855–864

    2016

  9. [17]

    Jensen, and Bin Yang

    Chenjuan Guo, Christian S. Jensen, and Bin Yang. 2014. Towards total traffic awareness. SIGMOD Record 43, 3 (2014), 18–23

    2014

  10. [18]

    Chenjuan Guo, Ronghui Xu, Bin Yang, Ye Yuan, Tung Kieu, Yan Zhao, and Christian S Jensen. 2024. Efficient stochastic routing in path-centric uncertain road networks. PVLDB 17, 11 (2024), 2893–2905

    2024

  11. [19]

    Chenjuan Guo, Bin Yang, Jilin Hu, and Christian Jensen. 2018. Learning to route with sparse trajectory sets. In ICDE. IEEE, 1073–1084

    2018

  12. [20]

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2016. Deep residual learning for image recognition. In CVPR. 770–778

    2016

  13. [21]

    Jiawei Jiang, Dayan Pan, Houxing Ren, Xiaohan Jiang, Chao Li, and Jingyuan Wang. 2023. Self-supervised trajectory representation learning with temporal Conference acronym ’XX, June 03–05, 2025, Woodstock, NY Ronghui Xu et al. regularities and travel semantics. In ICDE. IEEE

    2023

  14. [22]

    Jacob Devlin Ming-Wei Chang Kenton and Lee Kristina Toutanova. 2019. BERT: Pre-training of deep bidirectional transformers for language understanding. In NAACL-HLT. 4171–4186

    2019

  15. [23]

    Jensen, and Bac Le

    Duc Kieu, Tung Kieu, Peng Han, Bin Yang, Christian S. Jensen, and Bac Le. 2024. TEAM: Topological evolution-aware framework for traffic forecasting.PVLDB 18 (2024)

    2024

  16. [24]

    Zhonghang Li, Lianghao Xia, Jiabin Tang, Yong Xu, Lei Shi, Long Xia, Dawei Yin, and Chao Huang. 2024. Urbangpt: Spatio-temporal large language models. In KDD. 5351–5362

    2024

  17. [25]

    Zhonghang Li, Lianghao Xia, Yong Xu, and Chao Huang. 2024. FlashST: A Simple and Universal Prompt-Tuning Framework for Traffic Prediction. InICML

    2024

  18. [26]

    Paul Pu Liang, Amir Zadeh, and Louis-Philippe Morency. 2024. Foundations & trends in multimodal machine learning: Principles, challenges, and open questions. ACM Comput. Surv. 56, 10 (2024), 1–42

    2024

  19. [27]

    Ziqiao Liu, Hao Miao, Yan Zhao, Chenxi Liu, Kai Zheng, and Huan Li. 2024. LightTR: A lightweight framework for federated trajectory recovery. arXiv preprint arXiv:2405.03409 (2024)

    2024 arXiv

  20. [28]

    Yandi Lun, Hao Miao, Jiaxing Shen, Renzhi Wang, Xiang Wang, and Senzhang Wang. 2024. Resisting tul attack: balancing data privacy and utility on trajectory via collaborative adversarial learning. GeoInformatica 28, 3 (2024), 381–401

    2024

  21. [29]

    Zhipeng Ma, Zheyan Tu, Xinhai Chen, Yan Zhang, Deguo Xia, Guyue Zhou, Yilun Chen, Yu Zheng, and Jiangtao Gong. 2024. More than routing: Joint GPS and route modeling for refine trajectory representation learning. InWWW. 3064–3075

    2024

  22. [30]

    Sachit Menon and Carl V ondrick. 2022. Visual classification via description from large language models. In ICLR

    2022

  23. [31]

    Paul Newson and John Krumm. 2009. Hidden markov map matching through noise and sparseness. In SIGSPATIAL. 336–343

    2009

  24. [32]

    Zhicheng Pan, Yihang Wang, Yingying Zhang, Sean Bin Yang, Yunyao Cheng, Peng Chen, Chenjuan Guo, Qingsong Wen, Xiduo Tian, Yunliang Dou, et al. 2023. Magicscaler: Uncertainty-aware, predictive autoscaling. PVLDB 16, 12 (2023), 3808–3821

    2023

  25. [33]

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gre- gory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, et al

  26. [34]

    Simon Aagaard Pedersen, Bin Yang, and Christian S. Jensen. 2020. Anytime stochastic routing with hybrid learning. PVLDB 13, 9 (2020), 1555–1567

    2020

  27. [35]

    Jensen, and Jesper Møller

    Simon Aagaard Pedersen, Bin Yang, Christian S. Jensen, and Jesper Møller. 2023. Stochastic routing with arrival windows. ACM Trans. Spatial Algorithms Syst. 9, 4 (2023), 30:1–30:48

    2023

  28. [36]

    Shraman Pramanick, Yale Song, Sayan Nag, Kevin Qinghong Lin, Hardik Shah, Mike Zheng Shou, Rama Chellappa, and Pengchuan Zhang. 2023. EgoVLPv2: Egocentric video-language pre-training with fusion in the backbone. In ICCV. 5285–5297

    2023

  29. [37]

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. 2021. Learning transferable visual models from natural language supervision. In ICML. 8748–8763

    2021

  30. [38]

    Yi Tay, Mostafa Dehghani, Vinh Q Tran, Xavier Garcia, Jason Wei, Xuezhi Wang, Hyung Won Chung, Dara Bahri, Tal Schuster, Steven Zheng, et al. 2022. UL2: Unifying language learning paradigms. In ICLR

    2022

  31. [39]

    Romal Thoppilan, Daniel De Freitas, Jamie Hall, Noam Shazeer, Apoorv Kul- shreshtha, Heng-Tze Cheng, Alicia Jin, Taylor Bos, Leslie Baker, Yu Du, et al

  32. [40]

    Wenhui Wang, Hangbo Bao, Li Dong, Johan Bjorck, Zhiliang Peng, Qiang Liu, Kriti Aggarwal, Owais Khan Mohammed, Saksham Singhal, Subhojit Som, et al

  33. [41]

    Zhecheng Wang, Haoyuan Li, and Ram Rajagopal. 2020. Urban2vec: Incorporat- ing street view imagery and pois for multi-modal urban neighborhood embedding. In AAAI, V ol. 34. 1013–1020

    2020

  34. [42]

    Ronghui Xu, Weiming Huang, Jun Zhao, Meng Chen, and Liqiang Nie. 2023. A spatial and adversarial representation learning approach for land use classification with POIs. ACM Trans. Intell. Syst. Technol. 14, 6, Article 114 (nov 2023), 25 pages

    2023

  35. [43]

    Bin Yang, Jian Dai, Chenjuan Guo, Christian S Jensen, and Jilin Hu. 2018. PACE: a PAth-CEntric paradigm for stochastic path finding.VLDB J. 27 (2018), 153–178

    2018

  36. [44]

    Sean Bin Yang, Chenjuan Guo, and Bin Yang. 2020. Context-aware path ranking in road networks. IEEE Trans. Knowl. Data Eng. 34, 7 (2020), 3153–3168

    2020

  37. [45]

    Sean Bin Yang, Jilin Hu, Chenjuan Guo, Bin Yang, and Christian S. Jensen. 2023. LightPath: Lightweight and scalable path representation learning. In KDD. ACM, 2999–3010

    2023

  38. [46]

    Sean Bin Yang and Bin Yang. 2020. Learning to rank paths in spatial networks. In ICDE. IEEE, 2006–2009

    2020

  39. [47]

    Ziyi Yang, Yuwei Fang, Chenguang Zhu, Reid Pryzant, Dongdong Chen, Yu Shi, Yichong Xu, Yao Qian, Mei Gao, Yi-Ling Chen, et al. 2023. i-code: An integrative and composable multimodal learning framework. InAAAI, V ol. 37. 10880–10890

    2023

  40. [48]

    Yuan Yuan, Jingtao Ding, Jie Feng, Depeng Jin, and Yong Li. 2024. UniST: A prompt-empowered universal model for urban spatio-temporal prediction. InKDD. 4095–4106

    2024

  41. [49]

    Sen Zhang, Senzhang Wang, Xiang Wang, Shigeng Zhang, Hao Miao, and Junxing Zhu. 2022. Multi-task adversarial learning for semi-supervised trajectory-user linking. In ECML PKDD. Springer, 418–434

    2022

  42. [50]

    Kai Zhao, Chenjuan Guo, Yunyao Cheng, Peng Han, Miao Zhang, and Bin Yang

  43. [51]

    Yuanshao Zhu, Yongchao Ye, Shiyao Zhang, Xiangyu Zhao, and James Yu. 2023. Difftraj: Generating gps trajectory with diffusion probabilistic model. NeurIPS 36 (2023), 65168–65188

    2023

  44. [52]

    Yuanshao Zhu, James Jianqiao Yu, Xiangyu Zhao, Qidong Liu, Yongchao Ye, Wei Chen, Zijian Zhang, Xuetao Wei, and Yuxuan Liang. 2024. Controltraj: Controllable trajectory generation with topology-constrained diffusion model. In KDD. 4676–4687

    2024

  45. [54]

    PVLDB 17, 4 (2023), 753–765

    Multiple time series forecasting with dynamic graph modeling. PVLDB 17, 4 (2023), 753–765

    2023

  46. [2019]

    Pytorch: An imperative style, high-performance deep learning library

  47. [2022]

    arXiv preprint arXiv:2201.08239 (2022)

    Lamda: Language models for dialog applications. arXiv preprint arXiv:2201.08239 (2022)

    2022 arXiv

  48. [2023]

    Image as a foreign language: Beit pretraining for vision and vision-language tasks. In CVPR. 19175–19186

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.