Pith. sign in

REVIEW 5 major objections 5 minor 2 cited by

CAMS: A CityGPT-Powered Agentic Framework for Urban Human Mobility Simulation

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

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

Pith's one-line read CAMS claims that an agentic framework built on an urban-knowledgeable LLM can generate realistic daily trajectories in a real city from only a user profile, without external geospatial data.

desk verdict A well-built LLM-agent pipeline for mobility simulation whose central zero-shot claim is currently undercut by ambiguous test/train separation in the reflection and DPO stages. read the letter →

arxiv 2506.13599 v1 pith:2EYIKDNT submitted 2025-06-16 cs.CL cs.AI

Yuwei Du , Jie Feng , Jian Yuan , Yong Li This is my paper · ORCID
classification cs.CLcs.AI
keywords humanmobilitysimulationLLMagentsCityGPTurbanspatialknowledgetrajectorygenerationdirectpreferenceoptimizationagenticframeworkpatternextraction
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 proposes CAMS, an agentic framework for human mobility simulation that combines a large language model pre-trained on urban spatial knowledge with three cooperating modules. It claims that, starting from a short user profile such as age, income, occupation, and education, CAMS can generate a full day of realistic movement through a real city, including plausible home and work locations and semantically sensible activity sequences. This matters because traditional trajectory simulators need large travel-diary datasets or external map services, while CAMS aims to produce synthetic populations cheaply and interpretably from profiles alone. Experiments on two real-world datasets show CAMS outperforming existing deep-learning and LLM-based baselines on most metrics, especially spatial distribution and daily routine measures.

What carries the argument

The central object is a three-module agentic pipeline built on CityGPT, an LLM post-trained on urban spatial knowledge expressed in language. Mobility patterns flow from MobExtractor to GeoGenerator to TrajEnhancer, with urban structure encoded as a hierarchical address system (administrative area to subdistrict to street to point of interest). The load-bearing mechanism is the macro-to-micro cascaded generator for anchors, which uses reasoning-execution-reflection cycles and compares generated home and work locations against the collective ground-truth distribution, together with DPO training that progressively reinforces spatiotemporal continuity in generated trajectories.

What would settle it

Re-run the Anchor Location Extractor with the reflection stage restricted to training-user home and work distributions only, then compare the resulting Distance, Radius, and SD JSDs to the values in Table 2; if the gap shrinks materially, the reported zero-shot generalization was partly built on peeking at evaluation users.

Watch

Extended reading notes

Core claim

The paper contends that human mobility simulation can be recast as a language-driven generation problem, and that an LLM endowed with urban geospatial knowledge can carry the whole pipeline. CAMS decomposes the task into three stages: MobExtractor condenses raw trajectories of a small set of template users into natural-language mobility patterns and synthesizes new patterns for unseen users; GeoGenerator places each user's home and workplace through a macro-to-micro cascaded generator that reflects against the collective spatial distribution, then maps the rest of the day onto real points of interest; TrajEnhancer generates the final trajectory and aligns it with real trajectories through iterated direct preference optimization. The paper claims this yields superior performance without relying on externally provided geospatial information at generation time, and that the framework establishes a new paradigm for integrating urban-knowledgeable LLMs with agentic reasoning in mobility simulation.

Load-bearing premise

The load-bearing premise is that the ground-truth spatial distribution used in GeoGenerator's reflection stage is computed only from training users; the paper never states that, and the zero-shot claim collapses if evaluation users helped shape it.

Editorial extensions

If this is right

  • A city planner could generate synthetic daily trajectories for hypothetical residents defined only by profiles, without collecting travel surveys, and use them to test interventions like 15-minute-city layouts or transport policies.
  • Because new-user trajectories are generated without retraining, the framework could scale profile-to-trajectory simulation to large synthetic populations.
  • Generated trajectories come out as interpretable toponyms in a hierarchical address system, making them auditable against known urban facts rather than opaque coordinate sequences.
  • Repeated DPO iterations progressively reduce spatial-temporal discontinuity, so the same framework can be continuously refined as more real trajectories become available.

Reading between the lines

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

  • Not tested in the paper: withholding evaluation users' home and work coordinates from the collective distribution used in the reflection stage would directly test whether the reported spatial-metric advantages come from genuine zero-shot generalization or from fitting the test set.
  • The modular split between pattern extraction and geometry generation suggests the pipeline could transfer to other spatiotemporal simulation tasks, such as epidemic contact networks or retail footfall, by swapping the urban geometry source.
  • Because the reflection stage pushes individual outputs toward aggregate distributions, the framework may underrepresent rare or extreme mobility profiles; a subgroup analysis by occupation and income would quantify this flattening.
  • The paper's experiments use open map data and Foursquare points of interest to align the dataset, so the 'no external geospatial information' claim should be read as applying at query time, not to the model's training and alignment data.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper proposes CAMS, an agentic framework for urban human mobility simulation built on the urban LLM CityGPT. The framework combines three modules: MobExtractor, which extracts and synthesizes linguistic mobility patterns from user profiles; GeoGenerator, which generates home/work anchor points through a macro-to-micro cascaded reasoning loop with collective-distribution reflection and maps activity sequences to real urban locations using an enhanced CityGPT; and TrajEnhancer, which applies iterative DPO training to align generated trajectories with real trajectories. Experiments on Tencent and ChinaMobile Beijing datasets compare CAMS against mechanistic, deep-learning, and LLM baselines using JSD-based individual, collective, semantic, and hallucination metrics, and report that CAMS achieves the highest CMRR on both trajectory generation and recovery. The abstract claims that CAMS achieves superior performance without relying on externally provided geospatial information and that it establishes a new paradigm for integrating agentic frameworks with urban-knowledgeable LLMs.

Significance. If the central claim holds, CAMS would be a meaningful advance in LLM-based mobility simulation: it explicitly models hierarchical urban structure, uses collective distribution feedback to condition individual generation, and applies DPO to improve spatiotemporal continuity. The paper also provides a broad set of ablations and compares multiple LLM backbones, which is useful for the community. However, the empirical support is currently undermined by unresolved questions about test-set information leakage through the collective-distribution reflection and the DPO training data, and by the small, density-biased evaluation sample. These issues bear directly on the zero-shot generalization claim and must be resolved before the main conclusion can be accepted.

major comments (5)
  1. [Section 2.2.1 and Appendix A.1] The reflection stage of the Anchor Location Extractor adjusts generated home and workplace assignments by comparing against the 'ground-truth distribution,' but the paper never states that this distribution is computed exclusively from training users. Appendix A.1 says the collective-knowledge experiment uses a 44,313-user subset and that 100 recovery and 50 generation users are subsequently selected from the trajectory-dense users; the ground-truth home/work distribution may therefore include the evaluation users themselves. If it does, the reflection loop is directly fitting the test users' anchor-point distribution, and the reported spatial JSD values in Table 2 (e.g., Tencent Radius 0.0517 and SD 0.0457; ChinaMobile Radius 0.0507 and SD 0.0473) are inflated and do not support the zero-shot generalization claim. Please state explicitly whether the ground-truth distribution excludes all 150 evaluation users, and if it does not, rerun the evaluation using a distribution computed only from training users.
  2. [Section 2.3 and Appendix A.1] The DPO training data are described as constructed from 'the corpus output by CAMS and corresponding individuals' real trajectories,' with '300 training users and 1200 test users' reported in Appendix A.1, but this is never reconciled with the 100 recovery and 50 generation users used in Tables 2 and 3. If any of the 150 evaluation users' real trajectories serve as positive DPO examples, then TrajEnhancer is trained on test data and the comparison against the baselines is not a fair zero-shot test. Please define the exact disjoint sets of users used for DPO training, evaluation, and the collective distribution, and confirm that evaluation users are excluded from all positive DPO samples.
  3. [Section 3.1 and Appendix A.1] The evaluation uses only 100 recovery users and 50 generation users, selected as the top users by average daily trajectory points, and Tables 2 and 3 report single point estimates with no error bars, confidence intervals, or significance tests. Given the stochastic nature of LLM generation and the biased sampling toward data-rich users, the claimed 'superior performance on 11 out of 16 metrics' is not yet statistically supported. Please report repeated-generation variance and significance tests, and evaluate on a more representative sample or justify why the dense-trajectory subsample is sufficient for the generalization claim.
  4. [Section 2.2.2 and Section 3.1] The abstract and Section 3.2.1 state that CAMS achieves superior performance without relying on externally provided geospatial information, but GeoGenerator fine-tunes CityGPT on 10,000 POI question-answer pairs constructed from urban spatial data aligned to OSM and Foursquare, and Section 3.1 states that OSM road/AOI data and Foursquare POI data jointly represent urban space. Please clarify what counts as 'external' in this claim; as written, the claim is difficult to interpret because the method appears to use the same city's geospatial data during fine-tuning and evaluation, even if no map tool is queried at inference time.
  5. [Table 4] The Gemma3-27B row in Table 4 has identical values for Tencent and ChinaMobile (0.3994, 0.3903, 0.4160, 0.0695, 0.1717, 0.0506, 0.3265, 0.6570, 0.8252), while all other models differ across the two datasets, which strongly suggests a copy-paste error. This table supports the claim that enhanced CityGPT outperforms other LLM backbones, so the error must be corrected and the affected conclusions rechecked before the comparison can be interpreted.
minor comments (5)
  1. [Figure 1] Figure 1 contains garbled placeholder text ('?????????') and an anomalous '240 Duration 4h' fragment; the figure should be cleaned before publication.
  2. [Introduction and Tables 2-3] The name of the LLMob baseline is inconsistent: the introduction refers to 'LLM-Mob' while the tables and text use 'LLMob'; please unify the name and ensure the cited reference matches the method.
  3. [Section 3.1] The description of the deep-learning baseline setup says the training set is reduced to '3× test set,' but the exact number of users and trajectories is not specified; please state the concrete training sizes for each baseline.
  4. [Throughout] There are several typos, including 'accross' in Section 3.1 and 'it's ability' in Section 2.3; the manuscript should be proofread carefully.
  5. [Appendix A.1] The relationship between the 44,313-user subset, the top 1,500 users for the Individual Knowledge Extractor, the 300 training users and 1,200 test users for DPO, and the 150 evaluation users is not clearly described; presenting a single diagram of the data splits would remove ambiguity.

Circularity Check

1 steps flagged · score 6.0 of 10

Spatial-distribution reflection fits the evaluation target by construction, so the reported spatial gains are not an independent zero-shot prediction.

  1. fitted input called prediction [Section 2.2.1 'Reflection with collective distribution'; Appendix A.1 'Data preprocessing']
    "First, we transfer coordinates of all homes and workplaces into a hierarchical address representation ... Upon completing execution stage of all users, we compute spatial distribution of generated locations. Then, in reflection stage, the model does comparative analysis against ground-truth distribution and adjusts generation strategies for subsequent iterations. Finally, in execution stage, model dynamically adjust individual output to minimize distributional divergence. ..."

    The reflection stage explicitly takes the ground-truth home/work distribution as its target and 'dynamically adjust[s] individual output to minimize distributional divergence.' Appendix A.1 places the 50 generation users inside the same 44,313-user subset used for the Collective Knowledge Extractor, so that ground-truth distribution is computed from 'all homes and workplaces' and necessarily includes the evaluation users' own true anchors. The later evaluation compares generated trajectories against those same users' real trajectories on spatial metrics (Distance, Radius, SD, FVLoc). The low JSD values on those metrics are therefore a measure of how well CAMS reproduced a distribution it was directly optimized to match, not evidence of generalization from user profiles alone.

full rationale

The central spatial-generation claim is partially circular. Section 2.2.1's reflection mechanism is an explicit optimization loop whose input is the ground-truth home/work distribution of the dataset, and the evaluation users are a subset of the same 44,313 users used for that collective knowledge computation (Appendix A.1). Consequently, metrics such as Distance, Radius, SD, and FVLoc are partly self-comparisons: the generator was adjusted to minimize divergence from the very aggregate distribution that the evaluation measures against. This is a fitted input being reported as a predicted spatial distribution, which justifies a score of 6 rather than a clean 0-2. The DPO training description in Section 2.3 and Appendix A.1 also leaves the 300/1200-user DPO split unreconciled with the 150-user evaluation split, creating a separate leakage risk; however, the paper does not explicitly establish that evaluation users were used as positive DPO samples, so this is not counted as a demonstrated circular step. The self-citation of CityGPT is not scored as circular: CityGPT is compared against other LLMs in Table 4 and its geospatial knowledge is grounded in externally sourced OSM/Foursquare data, so the self-citation is not load-bearing in a definitional sense. Apart from the reflection-on-ground-truth issue, the rest of the framework (pattern extraction, similarity-based generation, fine-tuned urban mapping) has independent content and is evaluated against external baselines, so the circularity is partial rather than total.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central claim rests on a handful of domain assumptions about CityGPT's knowledge and the legitimacy of the collective distribution reflection. The free parameters are experimental choices that are not fully disclosed, which further limits independent verification.

free parameters (6)
  • POI retrieval radius = unknown; determined by average jump distance and mobility pattern
    Section 2.2.2 states the radius is determined by average jump distance between consecutive trajectory points and correlated with user mobility pattern, but no exact formula or value is given.
  • DPO negative-sample threshold = 5 (quality score)
    Appendix A.1: textual outputs scoring above 5 are selected as negative samples; threshold is an arbitrary choice affecting DPO training.
  • Number of similar users K = not reported
    MobExtractor retrieves top K similar template users; K is not specified, affecting pattern generation.
  • CityGPT fine-tuning dataset size = 10,000 QA pairs
    Section 2.2.2 constructs 10,000 question-answer pairs; this hand-built dataset is a free input to the pipeline.
  • DPO iterations = 3 shown in figures
    Figure 4 shows CAMS-DPO-3; the optimal number is chosen empirically without a stated criterion.
  • User selection counts = top 150 and top 1,500
    Appendix A.1: only the top 150 and top 1,500 users by trajectory density are used; sample selection is arbitrary.
assumptions (4)
  • domain assumption CityGPT possesses accurate, transferable geospatial knowledge of Beijing
    The whole GeoGenerator relies on CityGPT's internalized urban knowledge to produce locations; if CityGPT's knowledge is incomplete or biased, generated trajectories will be wrong.
  • domain assumption Mobility patterns can be compressed into text and reconstructed faithfully by an LLM
    MobExtractor's compression-recovery assumes the LLM can distill raw trajectories into linguistic patterns and invert them; no guarantee this preserves the statistical properties needed for simulation.
  • ad hoc to paper Ground-truth collective distribution used in reflection is computed from training users only
    Section 2.2.1 compares generated distributions to ground truth without specifying the split; the validity of the evaluation depends on this unstated assumption.
  • domain assumption Trajectory data (Tencent, ChinaMobile) is representative of general urban mobility
    The evaluation generalizes from two Beijing datasets; if these datasets are biased, the comparative results may not transfer.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CAMS: A CityGPT-Powered Agentic Framework for Urban Human Mobility Simulation." pith.science (2026). https://pith.science/paper/2EYIKDNT

@misc{pith2026250613599,
  author       = {Pith},
  title        = {Pith review of: CAMS: A CityGPT-Powered Agentic Framework for Urban Human Mobility Simulation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2EYIKDNT}},
  note         = {Machine review of arXiv:2506.13599}
}
read the original abstract

Human mobility simulation plays a crucial role in various real-world applications. Recently, to address the limitations of traditional data-driven approaches, researchers have explored leveraging the commonsense knowledge and reasoning capabilities of large language models (LLMs) to accelerate human mobility simulation. However, these methods suffer from several critical shortcomings, including inadequate modeling of urban spaces and poor integration with both individual mobility patterns and collective mobility distributions. To address these challenges, we propose \textbf{C}ityGPT-Powered \textbf{A}gentic framework for \textbf{M}obility \textbf{S}imulation (\textbf{CAMS}), an agentic framework that leverages the language based urban foundation model to simulate human mobility in urban space. \textbf{CAMS} comprises three core modules, including MobExtractor to extract template mobility patterns and synthesize new ones based on user profiles, GeoGenerator to generate anchor points considering collective knowledge and generate candidate urban geospatial knowledge using an enhanced version of CityGPT, TrajEnhancer to retrieve spatial knowledge based on mobility patterns and generate trajectories with real trajectory preference alignment via DPO. Experiments on real-world datasets show that \textbf{CAMS} achieves superior performance without relying on externally provided geospatial information. Moreover, by holistically modeling both individual mobility patterns and collective mobility constraints, \textbf{CAMS} generates more realistic and plausible trajectories. In general, \textbf{CAMS} establishes a new paradigm that integrates the agentic framework with urban-knowledgeable LLMs for human mobility simulation.

Figures

Figures reproduced from arXiv: 2506.13599 by the authors.

Figure 1
Figure 1. CAMS: CityGPT-powered Agentic framework for urban human Mobility Simulation, including MobExtractor, GeoGenerator and TrajEnhancer. understanding urban space. As a result, studies such as CoPB [31] and LLM-Mob [37] have proposed specific mechanisms within their frameworks to mitigate this limitation and harness the strengths of LLMs for sequential modeling and reasoning. However, these approaches typically combine s… view at source ↗
Figure 2
Figure 2. Methodological comparisons in Ur￾banMapper using Tencent dataset ActProb Distance SI SD STVD 0.0 0.1 0.2 0.3 0.4 0.5 0.6 JSD Value CAMS-E w/o C w/o I w/o I&C [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 4
Figure 4. DPO result analysis of TrajEnhancer. of generated trajectories to approximate real-world mobility patterns. Variations of each metric are visualized in Figure 4a. Comparison of different methodologies in UrbanMapper. By comparing results of using CityGPT￾enhanced(CAMS-E), map tools(CAMS-M) and social networks(CAMS-S) in [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figures from the paper (1 more)
Figure 5
Figure 5. Figure 5: Comparison of daily move￾ment patterns [PITH_FULL_IMAGE:figures/full_fig_p015_5.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

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

  1. Mobility-Aware Cache Framework for Scalable LLM-Based Human Mobility Simulation

    cs.AI 2026-02 conditional novelty 6.0 of 10

    A latent-space reasoning cache with a lightweight decoder cuts the cost of LLM-based human mobility simulation by roughly 40-90% while keeping trajectory quality comparable.

  2. Bridging Individual and Collective Realism in LLM-Based Human Mobility Simulation via Mobility Scaling-Law Guidance

    cs.MA 2026-02 conditional novelty 6.0 of 10

    M2LSimu uses population-level mobility statistics as a reward signal to iteratively adjust LLM prompts, improving simulated trajectories' match to real mobility patterns.

Reference graph

Works this paper leans on

45 extracted references · 21 canonical work pages · cited by 2 Pith papers

  1. [1]

    Lamp: A language model on the map

    Pasquale Balsebre, Weiming Huang, and Gao Cong. Lamp: A language model on the map. arXiv preprint arXiv:2403.09059, 2024

    arXiv 2024

  2. [2]

    Large language models are zero-shot next location predictors

    Ciro Beneduce, Bruno Lepri, and Massimiliano Luca. Large language models are zero-shot next location predictors. arXiv preprint arXiv:2405.20962, 2024

    arXiv 2024

  3. [3]

    Are large language models geospatially knowledgeable? In Proceedings of the 31st ACM International Conference on Advances in Geographic Information Systems , pages 1–4, 2023

    Prabin Bhandari, Antonios Anastasopoulos, and Dieter Pfoser. Are large language models geospatially knowledgeable? In Proceedings of the 31st ACM International Conference on Advances in Geographic Information Systems , pages 1–4, 2023

    work page 2023

  4. [4]

    The scaling laws of human travel

    Dirk Brockmann, Lars Hufnagel, and Theo Geisel. The scaling laws of human travel. Nature, 439(7075):462–465, 2006

    work page 2006

  5. [5]

    Language models are few-shot learners

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learners. Advances in neural information processing systems , 33:1877–1901, 2020

    1901

  6. [6]

    Elements of urban form

    Nicola Dempsey, Caroline Brown, Shibu Raman, Sergio Porta, Mike Jenks, Colin Jones, and Glen Bramley. Elements of urban form. Dimensions of the sustainable city , pages 21–51, 2010

    work page 2010

  7. [7]

    Understanding world or predicting future? a comprehensive survey of world models

    Jingtao Ding, Yunke Zhang, Yu Shang, Yuheng Zhang, Zefang Zong, Jie Feng, Yuan Yuan, Hongyuan Su, Nian Li, Nicholas Sukiennik, et al. Understanding world or predicting future? a comprehensive survey of world models. arXiv preprint arXiv:2411.14499, 2024

    arXiv 2024

  8. [8]

    Trajagent: An agent framework for unified trajectory modelling

    Yuwei Du, Jie Feng, Jie Zhao, and Yong Li. Trajagent: An agent framework for unified trajectory modelling. arXiv preprint arXiv:2410.20445, 2024

    arXiv 2024

Show all 45 references
  1. [9]

    Learning to simulate human mobility

    Jie Feng, Zeyu Yang, Fengli Xu, Haisu Yu, Mudan Wang, and Yong Li. Learning to simulate human mobility. In Proceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining , pages 3426–3433, 2020

    2020

  2. [10]

    Citygpt: Empowering urban spatial cognition of large language models

    Jie Feng, Tianhui Liu, Yuwei Du, Siqi Guo, Yuming Lin, and Yong Li. Citygpt: Empowering urban spatial cognition of large language models. arXiv preprint arXiv:2406.13948, 2024

    2024 arXiv

  3. [11]

    Agentmove: A large language model based agentic framework for zero-shot next location prediction

    Jie Feng, Yuwei Du, Jie Zhao, and Yong Li. Agentmove: A large language model based agentic framework for zero-shot next location prediction. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Lan...

    2025

  4. [12]

    S3: Social-network simulation system with large language model-empowered agents

    Chen Gao, Xiaochong Lan, Zhihong Lu, Jinzhu Mao, Jinghua Piao, Huandong Wang, Depeng Jin, and Yong Li. S3: Social-network simulation system with large language model-empowered agents. arXiv preprint arXiv:2307.14984, 2023. 10

    2023 arXiv

  5. [13]

    Large language models empowered agent-based modeling and simulation: A survey and perspectives

    Chen Gao, Xiaochong Lan, Nian Li, Yuan Yuan, Jingtao Ding, Zhilun Zhou, Fengli Xu, and Yong Li. Large language models empowered agent-based modeling and simulation: A survey and perspectives. Humanities and Social Sciences Communications , 11(1):1–24, 2024

    2024

  6. [14]

    Mobility-llm: Learning visiting intentions and travel preference from human mobility data with large language models

    Letian Gong, Yan Lin, Yiwen Lu, Xuedi Han, Yichen Liu, Shengnan Guo, Youfang Lin, Huaiyu Wan, et al. Mobility-llm: Learning visiting intentions and travel preference from human mobility data with large language models. Advances in Neural Information Processing Systems , 37: 36...

    2024

  7. [15]

    Language models represent space and time

    Wes Gurnee and Max Tegmark. Language models represent space and time. arXiv preprint arXiv:2310.02207, 2023

    2023 arXiv

  8. [16]

    The timegeo modeling framework for urban mobility without travel surveys

    Shan Jiang, Yingxiang Yang, Siddharth Gupta, Daniele Veneziano, Shounak Athavale, and Marta C González. The timegeo modeling framework for urban mobility without travel surveys. Proceedings of the National Academy of Sciences , 113(37):E5370–E5378, 2016

    2016

  9. [17]

    Towards full delegation: Designing ideal agentic behaviors for travel planning

    Song Jiang, Da JU, Andrew Cohen, Sasha Mitts, Aaron Foss, Justine T Kao, Xian Li, and Yuandong Tian. Towards full delegation: Designing ideal agentic behaviors for travel planning. arXiv preprint arXiv:2411.13904, 2024

    2024 arXiv

  10. [18]

    Be more real: Travel diary generation using llm agents and individual profiles

    Xuchuan Li, Fei Huang, Jianrong Lv, Zhixiong Xiao, Guolong Li, and Yang Yue. Be more real: Travel diary generation using llm agents and individual profiles. arXiv preprint arXiv:2407.18932, 2024

    2024 arXiv

  11. [19]

    The scaling of human mobility by taxis is exponential

    Xiao Liang, Xudong Zheng, Weifeng Lv, Tongyu Zhu, and Ke Xu. The scaling of human mobility by taxis is exponential. Physica A: Statistical Mechanics and its Applications , 391(5): 2135–2144, 2012

    2012

  12. [20]

    Act2loc: a synthetic trajec- tory generation method by combining machine learning and mechanistic models

    Kang Liu, Xin Jin, Shifen Cheng, Song Gao, Ling Yin, and Feng Lu. Act2loc: a synthetic trajec- tory generation method by combining machine learning and mechanistic models. International Journal of Geographical Information Science , 38(3):407–431, 2024

    2024

  13. [21]

    Practical synthetic human trajectories generation based on variational point processes

    Qingyue Long, Huandong Wang, Tong Li, Lisi Huang, Kun Wang, Qiong Wu, Guangyu Li, Yanping Liang, Li Yu, and Yong Li. Practical synthetic human trajectories generation based on variational point processes. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery ...

    2023

  14. [22]

    On the opportunities and challenges of foundation models for geospatial artificial intelligence

    Gengchen Mai, Weiming Huang, Jin Sun, Suhang Song, Deepak Mishra, Ninghao Liu, Song Gao, Tianming Liu, Gao Cong, Yingjie Hu, et al. On the opportunities and challenges of foundation models for geospatial artificial intelligence. arXiv preprint arXiv:2304.06798, 2023

    2023 arXiv

  15. [23]

    Geollm: Extracting geospatial knowledge from large language models

    Rohin Manvi, Samar Khanna, Gengchen Mai, Marshall Burke, David Lobell, and Stefano Ermon. Geollm: Extracting geospatial knowledge from large language models. arXiv preprint arXiv:2310.06213, 2023

    2023 arXiv

  16. [24]

    Large language models are geographically biased

    Rohin Manvi, Samar Khanna, Marshall Burke, David Lobell, and Stefano Ermon. Large language models are geographically biased. arXiv preprint arXiv:2402.02680, 2024

    2024 arXiv

  17. [25]

    Introducing chatgpt

    OpenAI. Introducing chatgpt. https://openai.com/blog/chatgpt/, 2022

    2022

  18. [26]

    Data-driven generation of spatio-temporal routines in human mobility

    Luca Pappalardo and Filippo Simini. Data-driven generation of spatio-temporal routines in human mobility. Data Mining and Knowledge Discovery , 32(3):787–829, 2018

    2018

  19. [27]

    Future directions in human mobility science

    Luca Pappalardo, Ed Manley, Vedran Sekara, and Laura Alessandretti. Future directions in human mobility science. Nature computational science, 3(7):588–600, 2023

    2023

  20. [28]

    Agentsociety: Large-scale simulation of llm- driven generative agents advances understanding of human behaviors and society.arXiv preprint arXiv:2502.08691, 2025

    Jinghua Piao, Yuwei Yan, Jun Zhang, Nian Li, Junbo Yan, Xiaochong Lan, Zhihong Lu, Zhiheng Zheng, Jing Yi Wang, Di Zhou, et al. Agentsociety: Large-scale simulation of llm- driven generative agents advances understanding of human behaviors and society.arXiv preprint arXiv:2502...

    2025 arXiv

  21. [29]

    Gpt4geo: How a language model sees the world’s geography

    Jonathan Roberts, Timo Lüddecke, Sowmen Das, Kai Han, and Samuel Albanie. Gpt4geo: How a language model sees the world’s geography. arXiv preprint arXiv:2306.00020, 2023. 11

    2023 arXiv

  22. [30]

    Structure of urban movements: polycentric activity and entangled hierarchical flows

    Camille Roth, Soong Moon Kang, Michael Batty, and Marc Barthélemy. Structure of urban movements: polycentric activity and entangled hierarchical flows. PloS one, 6(1):e15923, 2011

    2011

  23. [31]

    Beyond imitation: Generating human mobility from context-aware reasoning with large language models

    Chenyang Shao, Fengli Xu, Bingbing Fan, Jingtao Ding, Yuan Yuan, Meng Wang, and Yong Li. Beyond imitation: Generating human mobility from context-aware reasoning with large language models. arXiv e-prints, pages arXiv–2402, 2024

    2024

  24. [32]

    Llama 2: Open foundation and fine-tuned chat models

    Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, et al. Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288, 2023

    2023 arXiv

  25. [33]

    An extended exploration and preferential return model for human mobility simulation at individual and collective levels

    Jianying Wang, Lei Dong, Ximeng Cheng, Weijun Yang, and Yu Liu. An extended exploration and preferential return model for human mobility simulation at individual and collective levels. Physica A: Statistical Mechanics and Its Applications , 534:121921, 2019

    2019

  26. [34]

    Large language models as urban residents: An llm agent framework for personal mobility generation

    Jiawei Wang, Renhe Jiang, Chuang Yang, Zengqing Wu, Ryosuke Shibasaki, Noboru Koshizuka, Chuan Xiao, et al. Large language models as urban residents: An llm agent framework for personal mobility generation. Advances in Neural Information Processing Systems , 37:124547– 124574, 2024

    2024

  27. [35]

    A survey on large language model based autonomous agents

    Lei Wang, Chen Ma, Xueyang Feng, Zeyu Zhang, Hao Yang, Jingsen Zhang, Zhiyuan Chen, Jiakai Tang, Xu Chen, Yankai Lin, et al. A survey on large language model based autonomous agents. Frontiers of Computer Science, 18(6):186345, 2024

    2024

  28. [36]

    Where would i go next? large language models as human mobility predictors

    Xinglei Wang, Meng Fang, Zichao Zeng, and Tao Cheng. Where would i go next? large language models as human mobility predictors. arXiv preprint arXiv:2308.15197, 2023

    2023 arXiv

  29. [37]

    Simulating human- like daily activities with desire-driven autonomy

    Yiding Wang, Yuxuan Chen, Fangwei Zhong, Long Ma, and Yizhou Wang. Simulating human- like daily activities with desire-driven autonomy. arXiv preprint arXiv:2412.06435, 2024

    2024 arXiv

  30. [38]

    Emergent abilities of large language models

    Jason Wei, Yi Tay, Rishi Bommasani, Colin Raffel, Barret Zoph, Sebastian Borgeaud, Dani Yogatama, Maarten Bosma, Denny Zhou, Donald Metzler, et al. Emergent abilities of large language models. arXiv preprint arXiv:2206.07682, 2022

    2022 arXiv

  31. [39]

    Towards large reasoning models: A survey of reinforced reasoning with large language models

    Fengli Xu, Qianyue Hao, Zefang Zong, Jingwei Wang, Yunke Zhang, Jingyi Wang, Xiaochong Lan, Jiahui Gong, Tianjian Ouyang, Fanjin Meng, et al. Towards large reasoning models: A survey of reinforced reasoning with large language models. arXiv preprint arXiv:2501.09686, 2025

    2025 arXiv

  32. [40]

    Opencity: A scalable platform to simulate urban activities with massive llm agents

    Yuwei Yan, Qingbin Zeng, Zhiheng Zheng, Jingzhe Yuan, Jie Feng, Jun Zhang, Fengli Xu, and Yong Li. Opencity: A scalable platform to simulate urban activities with massive llm agents. arXiv preprint arXiv:2410.21286, 2024

    2024 arXiv

  33. [41]

    React: Synergizing reasoning and acting in language models

    Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, and Yuan Cao. React: Synergizing reasoning and acting in language models. In International Conference on Learning Representations (ICLR), 2023

    2023

  34. [42]

    Neeko: Leveraging dynamic lora for efficient multi-character role-playing agent

    Xiaoyan Yu, Tongxu Luo, Yifan Wei, Fangyu Lei, Yiming Huang, Hao Peng, and Liehuang Zhu. Neeko: Leveraging dynamic lora for efficient multi-character role-playing agent. arXiv preprint arXiv:2402.13717, 2024

    2024 arXiv

  35. [43]

    Activity trajectory generation via modeling spatiotemporal dynamics

    Yuan Yuan, Jingtao Ding, Huandong Wang, Depeng Jin, and Yong Li. Activity trajectory generation via modeling spatiotemporal dynamics. In Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages 4752–4762, 2022

    2022

  36. [44]

    Spatio-temporal diffusion point processes

    Yuan Yuan, Jingtao Ding, Chenyang Shao, Depeng Jin, and Yong Li. Spatio-temporal diffusion point processes. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pages 3173–3184, 2023

    2023

  37. [45]

    Spatial planning of urban communities via deep reinforcement learning

    Yu Zheng, Yuming Lin, Liang Zhao, Tinghai Wu, Depeng Jin, and Yong Li. Spatial planning of urban communities via deep reinforcement learning. Nature Computational Science, 3(9): 748–762, 2023. 12 A Appendix A.1 Data preprocessing Trajectory data preprocessing Among the 100,000...

    2023

Pith tools

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