Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

Foundation Models for Geospatial Reasoning: Assessing Capabilities of Large Language Models in Understanding Geometries and Topological Spatial Relations

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Large language models can infer topological spatial relations from WKT geometry text with over 60% accuracy, and GPT-4 few-shot reaches 66.6%, this paper reports.

desk verdict Useful exploratory benchmark for LLMs on DE-9IM with real WKT geometries, but the headline accuracy is tied to an underspecified hand-built test set and should not be treated as a stable estimate. read the letter →

arxiv 2505.17136 v1 pith:QOBG5UF3 submitted 2025-05-22 cs.CL cs.AI

Yuhan Ji , Song Gao , Ying Nie , Ivan Majić , Krzysztof Janowicz This is my paper
classification cs.CLcs.AI
keywords GeoAIgeospatialreasoningtopologicalspatialrelationsDE-9IMlargelanguagemodelswell-knowntextqueryexpansionconceptualneighborhood
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper asks whether a large language model can take two vector geometries written in WKT text and infer the topological relation between them, and whether that ability can support spatial queries. The authors build a real-world dataset of geometry pairs from Madison, Wisconsin, encode the geometries as text and as embeddings, and test GPT-3.5, GPT-4, and DeepSeek-R1 under several prompting schemes. They report that both embedding-based classifiers and GPT-based question answering exceed 0.6 average accuracy on seven DE-9IM predicates, with GPT-4 few-shot reaching 0.666. They also report that LLM-generated synthetic geometries, added to a spatial query, improve retrieval ranking of the target subject beyond direct queries, and that GPT-4 translates some vernacular place descriptions into formal predicates, though with instance-dependent success. If these results are right, natural-language spatial queries against vector data become plausible without dedicated spatial reasoning software.

What carries the argument

The load-bearing mechanism is the WKT textual encoding of geometries combined with the DE-9IM predicate set: point, line, and polygon geometries become plain strings, and the seven relations (equals, disjoint, crosses, touches, contains, within, overlaps) become the classification target. Two pipelines use that encoding: an embedding pipeline that feeds concatenated geometry embeddings into a random forest, and a prompting pipeline that presents the WKT strings to a generative model with zero-shot, few-shot, or chain-of-thought prompts. The evaluation is sharpened by the conceptual neighborhood graph and topological distance, which measure whether wrong answers are near-misses in a way that matches human and formal notions of relation similarity.

What would settle it

Sample a new set of real-world geometry pairs without the nearby-buffer and interpolation rules, run the same GPT-4 few-shot prompt, and see whether accuracy stays above 0.6; if it falls well below, the reported capability is tied to the artificial triplet construction.

Watch

Extended reading notes

Core claim

The central claim is that the well-known-text (WKT) string for a geometry retains enough geometric information for a large language model to reason about topology: when shown two WKT geometries, GPT-4 with few-shot examples picks the correct DE-9IM predicate about two-thirds of the time, and the embedding-plus-random-forest pipeline stays above 0.6. The paper further claims that this capability transfers to retrieval: generating a synthetic WKT geometry that satisfies a query and using it to expand the query embedding raises the mean reciprocal rank and Hits@K of the true subject, and reversing the predicate direction in object queries improves performance because it matches everyday language. Finally, the paper claims GPT-4 can convert some vernacular descriptions such as "is bordered by" into formal predicates, often landing in the conceptual neighborhood of the true relation when it does not hit exactly.

Load-bearing premise

The accuracy figures rest on a hand-constructed dataset in which "disjoint" pairs were drawn from a small nearby buffer and "equals" pairs were made by inserting extra points along lines, so if real geometry distributions differ, the 0.6-plus accuracy may not carry over.

Editorial extensions

If this is right

  • If the 0.666 GPT-4 few-shot result holds, an off-the-shelf language model can serve as a weak but real topological relation reasoner for WKT vector data, useful in question-answering and data-linking workflows.
  • Spatial query expansion with one LLM-generated geometry is a viable complement to direct text-embedding search, improving top-5 retrieval from 21.2% to 23.8% in this study.
  • Object retrieval improves when the query uses the reversed predicate, such as "within" instead of "contains", because it aligns with common spatial phrasing.
  • Chain-of-thought prompting does not help this task and can hurt, so reasoning about coordinates and topology should not be treated like standard text reasoning.
  • Adding place or geometry context in vernacular conversion sometimes helps and sometimes hurts; geometry-type context tends to help more than place-type context, but the effect is not uniform.

Reading between the lines

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

  • The accuracy ceiling near 0.66 suggests tokenization discards some metric detail; a geometry-aware tokenizer or fine-tuning on WKT could plausibly push past it, but that is an extension the paper leaves untested.
  • The dataset intentionally restricts "disjoint" pairs to nearby objects and creates "equals" by interpolating points, so the reported accuracy is probably optimistic for real GIS collections where far-apart disjoint pairs dominate and true equals are rare.
  • The conceptual-neighborhood analysis implies that a graded evaluation using topological distance is more informative than strict accuracy for spatial language, and future LLM spatial reasoning benchmarks could report both.
  • A testable next step is to combine the LLM-generated query expansion with a GIS engine that checks candidate geometries, rather than relying on embedding similarity alone, since the paper's retrieval numbers remain modest.
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 investigates the ability of large language models (GPT-3.5-turbo, GPT-4, DeepSeek-R1-14B) to reason about topological spatial relations from Well-Known Text (WKT) geometry representations. It introduces three evaluation tasks: (1) classifying DE-9IM topological predicates between two WKT geometries, (2) retrieving spatial entities from semantic queries, optionally expanded with LLM-generated synthetic geometries, and (3) converting vernacular language descriptions of spatial relations (from DBpedia) into formal DE-9IM predicates. The evaluation uses real-world geospatial datasets for Madison, Wisconsin, and DBpedia place descriptions, with ground-truth labels computed using GeoPandas. The main reported results are that embedding-based and prompt-based approaches with GPT models achieve over 0.6 average accuracy on Task 1, with GPT-4 few-shot reaching 0.666 (Section 5.1.2, Table 5), and that LLM-generated geometries improve retrieval performance in Task 2 (Section 5.2.3, Table 6). The paper also provides confusion analysis using conceptual neighborhoods and a discussion of context-dependent performance in Task 3.

Significance. If the results are robust, this paper offers a systematic, multi-task evaluation of LLMs for geospatial topological reasoning, with practical implications for GeoAI and spatial query systems. Strengths include the open availability of data and code, the use of GIS-computed ground truth rather than model-mediated labels, the comparison of three model families and multiple prompting strategies, and the analysis of errors in terms of conceptual neighborhoods. The paper also makes a useful contribution by showing that CoT prompting does not always help spatial reasoning and that LLM-generated geometries can aid semantic retrieval. However, the central quantitative claims rest on a hand-constructed benchmark whose sampling rules are only partially specified, and the per-cell test sets are small; these issues must be addressed before the reported accuracies can be interpreted as generalizable estimates of LLM performance on DE-9IM reasoning.

major comments (4)
  1. [Section 4.1.1, bullet (3)] The construction of 'disjoint' triplets is described as restricting subject geometries to lie 'within a smaller buffer of the objects' to avoid easy identification, but no buffer size, distance metric, or sampling algorithm is specified. Since 'disjoint' is applicable to every geometry-type combination and appears prominently in the average accuracy reported in Table 5, this undocumented parameter directly determines the difficulty of the benchmark. A tight buffer produces hard, near-touching disjoint pairs that depress accuracy; a generous buffer produces trivial far-apart pairs that inflate it. The headline 'over 0.6' and GPT-4 few-shot '0.666' numbers are therefore not reproducible properties of LLM spatial reasoning but of the specific, unreported sampling choice. Please specify the buffer size and the exact sampling procedure, and ideally provide a sensitivity analysis over buffer sizes to show how the headline accuracy varies.
  2. [Section 4.1.1 and Table 5] Each geometry-type/predicate combination is evaluated on only 40 test triplets, yet no confidence intervals, bootstrap estimates, or repeated-run variance are reported for the accuracy numbers in Table 5. With a test cell of 40 examples, a single misclassification changes accuracy by 2.5 percentage points, and the observed gaps between models (e.g., 0.666 for GPT-4 few-shot versus 0.633 for zero-shot, or 0.632 for the random forest) may not be statistically distinguishable. The paper should report uncertainty intervals (e.g., binomial confidence intervals) or increase the test sample size, and should avoid overinterpreting small differences in Table 5 as meaningful performance gains.
  3. [Section 4.1.1 (equals construction)] The 'equals' relation is synthesized by interpolating an additional 10% of points along lines or boundaries. This artificial transformation ensures that coordinate matching fails, but it is not clear that it represents how equal geometries occur in real geospatial datasets. The transformation may make the 'equals' classification easier (because the shape is preserved exactly) or harder (because the coordinate sets differ) than naturally occurring duplicates, and the reported average accuracy is therefore conditioned on this synthetic procedure. Please discuss the ecological validity of this construction, and if possible, also evaluate on naturally occurring equal geometries or describe why the synthetic approach is representative.
  4. [Section 5.1.1] DeepSeek-R1-14B results with few-shot prompting are omitted from the evaluation because the model 'largely ignored the provided examples.' While this is an acknowledged limitation, the paper does not report the actual validity values (format, geometry type, predicate) for those runs, making the cross-model comparison in Table 5 incomplete and not fully auditable. Please report the excluded numbers in an appendix or supplementary material, even if they are not used in the accuracy comparison, and clearly state that the current comparison is restricted to the prompting conditions each model could follow.
minor comments (6)
  1. [Abstract and Section 3.3] 'Well-known-text' should be 'Well-known text' (WKT) in both the abstract and the methodology text.
  2. [Section 5.1.2] 'GPT-4 with few-short promoting' is a typo and should be 'GPT-4 with few-shot prompting.'
  3. [Section 5.1.2] The sentence 'DeepSeek-R1-14B outperformed GPT-3.5-turbo (20B parameters)' is confusing: DeepSeek-R1-14B has 14B parameters, and GPT-3.5-turbo's parameter count is not publicly 20B; please revise to avoid a factual error.
  4. [Section 6.3] 'The mismatch between crispy geometry and the vague human perception' should use 'crisp geometry' rather than 'crispy.'
  5. [Table 5] The column label 'Dist(Incorrect)' is not defined in the table caption or Section 5.1.2; please state how the topological distance for incorrect predictions is aggregated.
  6. [Section 4.2 and Table 4] The prompt variant 'Zero-shot-dim' appears in Table 4 and Table 5 but is not defined in Section 4.2 or Section 3.4.1; please explain what this prompt variant does.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: ground-truth labels are computed externally by GeoPandas, and the reported accuracy numbers are not forced by construction.

full rationale

The paper's claims are empirical evaluations rather than analytical derivations. In Task 1, the ground-truth topological predicates are computed by GeoPandas from WKT geometries (Section 4.1.1), and the accuracies in Table 5 compare model outputs to those external labels. No parameter is fitted to the test labels, and the random forest is trained on a disjoint 160-triplet split and tested on 40 held-out triplets per cell, a standard train/test protocol. In Task 2, retrieval targets are ground-truth subjects from the spatial database; LLM-generated geometries only expand the query, and the MRR/Hits@K gains are measured against those external targets. In Task 3, the conversion target predicates are also computed by GIS tools from geometries, not by the model. The mildest self-referential element is that GPT-4 is used in Section 4.1.2 to extract the vernacular relation phrases from DBpedia abstracts that later form Task 3's test inputs; this can bias which phrases appear, but it does not define the ground-truth predicate and does not force the conversion accuracy. The self-citations (e.g., Ji and Gao 2023 in Sections 2.3 and 5.1.1) are background corroboration and are not load-bearing for the accuracy claims. The underspecified 'smaller buffer' for disjoint sampling and the 40-example test cells are reproducibility and generalizability concerns, not circularity. Overall, no significant circularity is present; the score of 2 reflects only minor self-referential data construction and background self-citations.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The paper does not fit mathematical constants; its free parameters are experimental design choices that control task difficulty. The axioms are mostly domain assumptions about the validity of WKT, GeoPandas ground truth, and the constructed datasets. No new physical or conceptual entities are introduced.

free parameters (5)
  • Disjoint sampling buffer = unspecified
    In Section 4.1.1, disjoint triplets are restricted to nearby entities within a smaller buffer to avoid trivial far-apart cases; the buffer size is not reported and directly changes the difficulty of the classification task.
  • Equals interpolation fraction = 10% for LineStrings; loop origin for Polygons
    In Section 4.1.1, equivalent LineString entities are created by interpolating 10% additional points, and Polygon equals by looping the origin point; this hand-chosen distortion level affects how hard it is to detect equals from coordinates.
  • Task 3 frequency threshold = at least 5 occurrences; sample 5 per combination
    Section 4.1.2 retains only context-conditioned conversion pairs occurring at least 5 times and samples 5 pairs per combination, which shapes the set of vernacular phrases evaluated.
  • Random forest tree count = 100
    Section 4.2.3 sets the estimator count to 100; this is a standard default and unlikely to be load-bearing, but it is a hand-chosen hyperparameter.
  • LLM temperature settings = 0 for GPT tasks, 0.6 for DeepSeek, 0.7 for geometry generation
    Section 4.2.3 sets temperatures to control determinism and creativity; these choices affect output distributions and therefore the reported accuracy.
assumptions (5)
  • domain assumption DE-9IM predicates with within/contains redefined as strict (not equals) and a modified decision tree (Section 3.2, Table 2) define the ground truth.
    The paper changes the standard OGC semantics to ensure mutual exclusivity. If a model was trained on standard OGC semantics where within includes equals, accuracy against this bespoke label scheme is not directly interpretable.
  • domain assumption WKT strings preserve the geometric information needed for topological reasoning when tokenized and embedded by LLMs (Section 3.3).
    The entire Task 1 and Task 2 pipeline assumes that converting vector geometry to WKT text does not lose the information needed for relation classification. This is not proven, and the results themselves suggest partial information loss.
  • domain assumption GeoPandas-computed topological relations on crisp administrative boundaries are the correct labels for vernacular place descriptions (Sections 3.4.3 and 6.3).
    The paper acknowledges that official boundaries can differ from human perception (e.g., the Shullsburg example), so the ground truth may disagree with the commonsense answer being tested.
  • ad hoc to paper The hand-constructed triplet sampling (200 per geometry-predicate combination, nearby disjoint, synthetic equals) yields a representative difficulty benchmark (Section 4.1.1).
    These sampling rules are specific to this study and there is no external benchmark to validate that they reflect real-world spatial relation distributions.
  • ad hoc to paper Rephrased DBpedia sentences are a valid proxy for authentic vernacular place descriptions (Section 7).
    The authors acknowledge they rewrote paragraphs from DBpedia, which loses anaphora and natural variation, limiting the external validity of Task 3 results.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Foundation Models for Geospatial Reasoning: Assessing Capabilities of Large Language Models in Understanding Geometries and Topological Spatial Relations." pith.science (2026). https://pith.science/paper/QOBG5UF3

@misc{pith2026250517136,
  author       = {Pith},
  title        = {Pith review of: Foundation Models for Geospatial Reasoning: Assessing Capabilities of Large Language Models in Understanding Geometries and Topological Spatial Relations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QOBG5UF3}},
  note         = {Machine review of arXiv:2505.17136}
}
read the original abstract

Applying AI foundation models directly to geospatial datasets remains challenging due to their limited ability to represent and reason with geographical entities, specifically vector-based geometries and natural language descriptions of complex spatial relations. To address these issues, we investigate the extent to which a well-known-text (WKT) representation of geometries and their spatial relations (e.g., topological predicates) are preserved during spatial reasoning when the geospatial vector data are passed to large language models (LLMs) including GPT-3.5-turbo, GPT-4, and DeepSeek-R1-14B. Our workflow employs three distinct approaches to complete the spatial reasoning tasks for comparison, i.e., geometry embedding-based, prompt engineering-based, and everyday language-based evaluation. Our experiment results demonstrate that both the embedding-based and prompt engineering-based approaches to geospatial question-answering tasks with GPT models can achieve an accuracy of over 0.6 on average for the identification of topological spatial relations between two geometries. Among the evaluated models, GPT-4 with few-shot prompting achieved the highest performance with over 0.66 accuracy on topological spatial relation inference. Additionally, GPT-based reasoner is capable of properly comprehending inverse topological spatial relations and including an LLM-generated geometry can enhance the effectiveness for geographic entity retrieval. GPT-4 also exhibits the ability to translate certain vernacular descriptions about places into formal topological relations, and adding the geometry-type or place-type context in prompts may improve inference accuracy, but it varies by instance. The performance of these spatial reasoning tasks offers valuable insights for the refinement of LLMs with geographical knowledge towards the development of geo-foundation models capable of geospatial reasoning.

Figures

Figures reproduced from arXiv: 2505.17136 by the authors.

Figure 1
Figure 1. The conceptual neighborhood of topological relations in RCC-8 (Randell et al., 1992) [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Overview of the workflow in this research. [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. The decision tree for the topological spatial relations. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: : The workflow for the topological spatial relation qualification task. [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: : Topological spatial relation qualification example inputs and outputs with different [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: : The workflow for the spatial query processing task. [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: : Prompt template used for geometry generation in the spatial query processing task. [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 8
Figure 8. Figure 8: : The workflow for the vernacular relations conversion task. [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]
Figure 9
Figure 9. Figure 9: : Prompt template used in the vernacular relation conversion task. [PITH_FULL_IMAGE:figures/full_fig_p014_9.png]
Figure 10
Figure 10. Figure 10: : Synthetic geometries generated by GPT-4 for LineString/LineString relations [PITH_FULL_IMAGE:figures/full_fig_p021_10.png]
Figure 11
Figure 11. Figure 11: : Confusion matrices between topological predicates in relation qualification. [PITH_FULL_IMAGE:figures/full_fig_p024_11.png]
Figure 12
Figure 12. Figure 12: : Confusion matrices between spatial predicates in geometry generation. [PITH_FULL_IMAGE:figures/full_fig_p025_12.png]
Figure 13
Figure 13. Figure 13: : Invalid synthetic geometries generated by GPT-4 for LineString/LineString relations [PITH_FULL_IMAGE:figures/full_fig_p026_13.png]

Discussion (0). Sign in 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. A Vision for Geo-Temporal Deep Research Systems: Towards Comprehensive, Transparent, and Reproducible Geo-Temporal Information Synthesis

    cs.CL 2025-06 conditional novelty 4.0 of 10

    A research agenda calling for geo-temporal reasoning in deep research systems, with no experiments or system implementation.

Reference graph

Works this paper leans on

118 extracted references · 52 canonical work pages · cited by 1 Pith paper

  1. [1]

    @esa (Ref

    \@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should not add it explicitly Type <Return> for now, but then later remove the command n...

  2. [2]

    \@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@first@sw \@firstoftwo \@ifundefined NAT@b*@#2 \@firstoftwo @num @NAT@ctr \@secondoft...

  3. [3]

    @open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibsetup #1 @NAT@ctr @ @openbib .11em \@plus.33em \@minus.07em 4000 4000 `\.\@m @bibit...

  4. [4]

    L., Almeida, D., Altenschmidt, J., Altman, S., Anadkat, S., et al

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

    arXiv 2023

  5. [5]

    Alayrac, J.-B., Donahue, J., Luc, P., Miech, A., Barr, I., Hasson, Y., Lenc, K., Mensch, A., Millican, K., Reynolds, M., et al. (2022). Flamingo: a visual language model for few-shot learning. Advances in Neural Information Processing Systems , 35:23716--23736

    2022

  6. [6]

    E., Barry, J., and Cohn, A

    Blackwell, R. E., Barry, J., and Cohn, A. G. (2024). Towards reproducible llm evaluation: Quantifying uncertainty in llm benchmark scores. arXiv preprint arXiv:2410.03492

    arXiv 2024

  7. [7]

    Bordes, A., Usunier, N., Garcia-Duran, A., Weston, J., and Yakhnenko, O. (2013). Translating embeddings for modeling multi-relational data. Advances in neural information processing systems , 26

    2013

  8. [8]

    Breiman, L. (2001). Random forests. Machine Learning , 45:5--32

    2001

Show all 118 references
  1. [9]

    D., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., et al

    Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J. D., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., et al. (2020). Language models are few-shot learners. Advances in Neural Information Processing Systems , 33:1877--1901

    2020

  2. [10]

    Carlson, L. A. and Logan, G. D. (2001). Using spatial terms to select an object. Memory & Cognition , 29(6):883--892

    2001

  3. [11]

    and Romano, G

    Carpineto, C. and Romano, G. (2012). A survey of automatic query expansion in information retrieval. ACM Computing Surveys (CSUR) , 44(1):1--50

    2012

  4. [12]

    Cervone, G., Sava, E., Huang, Q., Schnebele, E., Harrison, J., and Waters, N. (2016). Using Twitter for tasking remote-sensing data collection and damage assessment: 2013 boulder flood case study. International Journal of Remote Sensing , 37(1):100--124

    2016

  5. [13]

    Chen, H., Vasardani, M., Winter, S., and Tomko, M. (2018). A graph database model for knowledge extracted from place descriptions. ISPRS International Journal of Geo-Information , 7(6):221

    2018

  6. [14]

    G., Liu, D., Wang, S., Ouyang, J., and Yu, Q

    Chen, J., Cohn, A. G., Liu, D., Wang, S., Ouyang, J., and Yu, Q. (2015). A survey of qualitative spatial representations. The Knowledge Engineering Review , 30(1):106--136

    2015

  7. [15]

    Chen, M., Tworek, J., Jun, H., Yuan, Q., Pinto, H. P. D. O., Kaplan, J., Edwards, H., Burda, Y., Joseph, N., Brockman, G., et al. (2021). Evaluating large language models trained on code. arXiv preprint arXiv:2107.03374

    2021 arXiv

  8. [16]

    and Cohn, A

    Clementini, E. and Cohn, A. G. (2014). RCC*-9 and CBM . In International Conference on Geographic Information Science , pages 349--365. Springer

    2014

  9. [17]

    and Cohn, A

    Clementini, E. and Cohn, A. G. (2024). Extension of RCC*-9 to complex and three-dimensional features and its reasoning system. ISPRS International Journal of Geo-Information , 13(1):25

    2024

  10. [18]

    and Di Felice, P

    Clementini, E. and Di Felice, P. (1996). A model for representing topological relationships between complex geometric features in spatial databases. Information sciences , 90(1-4):121--136

    1996

  11. [19]

    Clementini, E., Di Felice, P., and Van Oosterom, P. (1993). A small set of formal topological relationships suitable for end-user interaction. In International symposium on spatial databases , pages 277--295. Springer

    1993

  12. [20]

    Clementini, E., Sharma, J., and Egenhofer, M. J. (1994). Modelling topological spatial relations: Strategies for query processing. Computers & graphics , 18(6):815--822

    1994

  13. [21]

    Cohn, A. G. (2023). An evaluation of ChatGPT-4's qualitative spatial reasoning capabilities in RCC-8 . arXiv preprint arXiv:2309.15577

    2023 arXiv

  14. [22]

    Cohn, A. G. and Blackwell, R. E. (2024a). Can large language models reason about the region connection calculus? arXiv preprint arXiv:2411.19589

    2024 arXiv

  15. [23]

    Cohn, A. G. and Blackwell, R. E. (2024b). Evaluating the Ability of Large Language Models to Reason About Cardinal Directions . In Adams, B., Griffin, A. L., Scheider, S., and McKenzie, G., editors, 16th International Conference on Spatial Information Theory (COSIT 2024) , vol...

    2024

  16. [24]

    Cohn, A. G. and Hazarika, S. M. (2001). Qualitative spatial representation and reasoning: An overview. Fundamenta informaticae , 46(1-2):1--29

    2001

  17. [25]

    Cohn, A. G. and Hernandez-Orallo, J. (2023). Dialectical language model evaluation: An initial appraisal of the commonsense spatial reasoning abilities of llms. arXiv preprint arXiv:2304.11164

    2023 arXiv

  18. [26]

    Cohn, A. G. and Renz, J. (2008). Qualitative spatial representation and reasoning. Foundations of Artificial Intelligence , 3:551--596

    2008

  19. [27]

    G., and Randell, D

    Cui, Z., Cohn, A. G., and Randell, D. A. (1993). Qualitative and topological relationships in spatial databases. In Advances in Spatial Databases: Third International Symposium, SSD'93 Singapore, June 23--25, 1993 Proceedings 3 , pages 296--315. Springer

    1993

  20. [28]

    Das, S. (2023). Evaluating the capabilities of large language models for spatial and situational understanding . PhD thesis, PhD thesis, Thesis (MA). University of Cambridge

    2023

  21. [29]

    Devlin, J., Chang, M.-W., Lee, K., and Toutanova, K. (2018). BERT : Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805

    2018 arXiv

  22. [30]

    Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., et al. (2020). An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929

    2020 arXiv

  23. [31]

    Du, S., Qin, Q., Chen, D., and Wang, L. (2005). Spatial data query based on natural language spatial relations. In Proceedings. 2005 IEEE International Geoscience and Remote Sensing Symposium, 2005. IGARSS'05. , volume 2, pages 1210--1213. IEEE

    2005

  24. [32]

    Egenhofer, M. J. and Al-Taha , K. K. (1992). Reasoning about gradual changes of topological relationships. In Goos, G., Hartmanis, J., Frank, A. U., Campari, I., and Formentini, U., editors, Theories and Methods of Spatio-Temporal Reasoning in Geographic Space , volume 639 of ...

    1992

  25. [33]

    Egenhofer, M. J. and Franzosa, R. D. (1991). Point-set topological spatial relations . International journal of geographical information systems , 5(2):161--174

    1991

  26. [34]

    Egenhofer, M. J. and Herring, J. R. (1991). Categorizing binary topological relations between regions, lines, and points in geographic databases. Technical report, Department of Surveying Engineering, University of Maine, Orono, Maine, USA

    1991

  27. [35]

    Egenhofer, M. J. and Mark, D. M. (1995). Modelling conceptual neighbourhoods of topological line-region relations. International journal of geographical information systems , 9(5):555--565

    1995

  28. [36]

    and Dube, S

    Fernandez, A. and Dube, S. (2023). Core building blocks: Next gen geo spatial gpt application. arXiv preprint arXiv:2310.11029

    2023 arXiv

  29. [37]

    Formica, A., Mazzei, M., Pourabbas, E., and Rafanelli, M. (2018). Approximate answering of queries involving polyline--polyline topological relationships. Information Visualization , 17(2):128--145

    2018

  30. [38]

    Frank, A. U. (1992). Qualitative spatial reasoning about distances and directions in geographic space. Journal of Visual Languages & Computing , 3(4):343--371

    1992

  31. [39]

    Freksa, C., Habel, C., and Wender, K. F. (1998). Spatial cognition: An interdisciplinary approach to representing and processing spatial knowledge , volume 1404. Springer Science & Business Media

    1998

  32. [40]

    Frolov, S., Hinz, T., Raue, F., Hees, J., and Dengel, A. (2021). Adversarial text-to-image synthesis: A review. Neural Networks , 144:187--209

    2021

  33. [41]

    Fulman, N., Memduho g lu, A., and Zipf, A. (2024). Distortions in judged spatial relations in large language models. The Professional Geographer , 76(6):703--711

    2024

  34. [42]

    and Goodchild, M

    Gao, S. and Goodchild, M. F. (2013). Asking spatial questions to identify GIS functionality. In 2013 Fourth International Conference on Computing for Geospatial Research and Application , pages 106--110. IEEE

    2013

  35. [43]

    Gao, S., Hu, Y., and Li, W. (2023). Handbook of Geospatial Artificial Intelligence . CRC Press

    2023

  36. [44]

    R., Hu, Y., Yang, J.-A., McKenzie, G., Ju, Y., Gong, L., Adams, B., and Yan, B

    Gao, S., Janowicz, K., Montello, D. R., Hu, Y., Yang, J.-A., McKenzie, G., Ju, Y., Gong, L., Adams, B., and Yan, B. (2017). A data-synthesis-driven method for detecting and extracting vague cognitive regions. International Journal of Geographical Information Science , 31(6):1245--1271

    2017

  37. [45]

    Gramacki, P., Martins, B., and Szyma \'n ski, P. (2024). Evaluation of code llms on geospatial code generation. In Proceedings of the 7th ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery , pages 54--62

    2024

  38. [46]

    Guo, D., Ge, S., Zhang, S., Gao, S., Tao, R., and Wang, Y. (2022). DeepSSN : A deep convolutional neural network to assess spatial scene similarity. Transactions in GIS , 26(4):1914--1938

    2022

  39. [47]

    Guo, D., Yang, D., Zhang, H., Song, J., Zhang, R., Xu, R., Zhu, Q., Ma, S., Wang, P., Bi, X., et al. (2025). Deepseek-r1: Incentivizing reasoning capability in LLMs via reinforcement learning . arXiv preprint arXiv:2501.12948

    2025 arXiv

  40. [48]

    Guo, R. (1998). Spatial objects and spatial relationships. Geo-spatial Information Science , 1(1):38--42

    1998

  41. [49]

    Head, C. G. (1984). The map as natural language: a paradigm for understanding. Cartographica: The International Journal for Geographic Information and Geovisualization , 21(1):1--32

    1984

  42. [50]

    Hu, Y., Goodchild, M., Zhu, A.-X., Yuan, M., Aydin, O., Bhaduri, B., Gao, S., Li, W., Lunga, D., and Newsam, S. (2024). A five-year milestone: reflections on advances and limitations in GeoAI research. Annals of GIS , 30(1):1--14

    2024

  43. [51]

    Hu, Y., Janowicz, K., and Prasad, S. (2014). Improving wikipedia-based place name disambiguation in short texts using structured data from dbpedia. In Proceedings of the 8th workshop on geographic information retrieval , pages 1--8

    2014

  44. [52]

    Hu, Y., Janowicz, K., Prasad, S., and Gao, S. (2015). Metadata topic harmonization and semantic search for linked-data-driven geoportals: A case study using ArcGIS Online . Transactions in GIS , 19(3):398--416

    2015

  45. [53]

    Z., and Joseph, K

    Hu, Y., Mai, G., Cundy, C., Choi, K., Lao, N., Liu, W., Lakhanpal, G., Zhou, R. Z., and Joseph, K. (2023). Geo-knowledge-guided gpt models improve the extraction of location descriptions from disaster-related social media messages. International Journal of Geographical Informa...

    2023

  46. [54]

    Huang, W., Wang, J., and Cong, G. (2024). Zero-shot urban function inference with street view images through prompting a pretrained vision-language model. International Journal of Geographical Information Science , 38(7):1414--1442

    2024

  47. [55]

    Jakubik, J., Roy, S., Phillips, C., Fraccaro, P., Godwin, D., Zadrozny, B., Szwarcman, D., Gomes, C., Nyirjesy, G., Edwards, B., et al. (2023). Foundation models for generalist geospatial artificial intelligence. arXiv preprint arXiv:2310.18660

    2023 arXiv

  48. [56]

    Janowicz, K. (2023). Philosophical foundations of GeoAI: Exploring sustainability, diversity, and bias in GeoAI and spatial data science . arXiv preprint arXiv:2304.06508 , pages 1--17

    2023 arXiv

  49. [57]

    Janowicz, K., Gao, S., McKenzie, G., Hu, Y., and Bhaduri, B. (2020). GeoAI : spatially explicit artificial intelligence techniques for geographic knowledge discovery and beyond. International Journal of Geographical Information Science , 34(4):625--636

    2020

  50. [58]

    A., and Hitzler, P

    Janowicz, K., Van Harmelen, F., Hendler, J. A., and Hitzler, P. (2015). Why the data train needs semantic rails. AI Magazine , 36(1):5--14

    2015

  51. [59]

    and Gao, S

    Ji, Y. and Gao, S. (2023). Evaluating the effectiveness of large language models in representing textual descriptions of geometry and spatial relations (short paper). In 12th International Conference on Geographic Information Science (GIScience 2023) , volume 43, pages 1--6. S...

    2023

  52. [60]

    B., Abdelmoty, A

    Jones, C. B., Abdelmoty, A. I., Finch, D., Fu, G., and Vaid, S. (2004). The spirit spatial search engine: Architecture, ontologies and spatial indexing. In Geographic Information Science: Third International Conference, GIScience 2004, Adelphi, MD, USA, October 20-23, 2004. Pr...

    2004

  53. [61]

    u chemann, S., Bannert, M., Dementieva, D., Fischer, F., Gasser, U., Groh, G., G \

    Kasneci, E., Se ler, K., K \"u chemann, S., Bannert, M., Dementieva, D., Fischer, F., Gasser, U., Groh, G., G \"u nnemann, S., H \"u llermeier, E., et al. (2023). ChatGPT for good? on opportunities and challenges of large language models for education. Learning and Individual ...

    2023

  54. [62]

    C., Lo, W.-Y., et al

    Kirillov, A., Mintun, E., Ravi, N., Mao, H., Rolland, C., Gustafson, L., Xiao, T., Whitehead, S., Berg, A. C., Lo, W.-Y., et al. (2023). Segment anything. arXiv preprint arXiv:2304.02643

    2023 arXiv

  55. [63]

    S., Reid, M., Matsuo, Y., and Iwasawa, Y

    Kojima, T., Gu, S. S., Reid, M., Matsuo, Y., and Iwasawa, Y. (2022). Large language models are zero-shot reasoners. Advances in neural information processing systems , 35:22199--22213

    2022

  56. [64]

    Kordjamshidi, P., Van Otterlo, M., and Moens, M.-F. (2011). Spatial role labeling: Towards extraction of spatial relations from natural language. ACM Transactions on Speech and Language Processing (TSLP) , 8(3):1--36

    2011

  57. [65]

    u ttler, H., Lewis, M., Yih, W.-t., Rockt \

    Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., K \"u ttler, H., Lewis, M., Yih, W.-t., Rockt \"a schel, T., et al. (2020). Retrieval-augmented generation for knowledge-intensive nlp tasks. Advances in Neural Information Processing Systems , 33:9459--9474

    2020

  58. [66]

    Li, W., Lee, H., Wang, S., Hsu, C.-Y., and Arundel, S. T. (2023). Assessment of a new GeoAI foundation model for flood inundation mapping. In Proceedings of the 6th ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery , pages 102--109

    2023

  59. [67]

    and Ning, H

    Li, Z. and Ning, H. (2023). Autonomous GIS : the next-generation AI -powered GIS . International Journal of Digital Earth , 16(2):4668--4686

    2023

  60. [68]

    and Lee, H

    Logeswaran, L. and Lee, H. (2018). An efficient framework for learning sentence representations. arXiv preprint arXiv:1803.02893

    2018 arXiv

  61. [69]

    Loglisci, C., Ienco, D., Roche, M., Teisseire, M., and Malerba, D. (2012). An unsupervised framework for topological relations extraction from geographic documents. In Database and Expert Systems Applications: 23rd International Conference, DEXA 2012, Vienna, Austria, Septembe...

    2012

  62. [70]

    D., and Jones, C

    M Hall, M., Smart, P. D., and Jones, C. B. (2011). Interpreting spatial language in image captions. Cognitive processing , 12:67--94

    2011

  63. [71]

    Mai, G., Cundy, C., Choi, K., Hu, Y., Lao, N., and Ermon, S. (2022a). Towards a foundation model for geospatial artificial intelligence (vision paper). In Proceedings of the 30th International Conference on Advances in Geographic Information Systems , pages 1--4

    2022

  64. [72]

    Mai, G., Huang, W., Sun, J., Song, S., Mishra, D., Liu, N., Gao, S., Liu, T., Cong, G., Hu, Y., et al. (2024). On the opportunities and challenges of foundation models for GeoAI (vision paper) . ACM Transactions on Spatial Algorithms and Systems , 10(2):1--46

    2024

  65. [73]

    Mai, G., Janowicz, K., Cai, L., Zhu, R., Regalia, B., Yan, B., Shi, M., and Lao, N. (2020). SE-KGE : A location-aware knowledge graph embedding model for geographic question answering and spatial semantic lifting. Transactions in GIS , 24(3):623--655

    2020

  66. [74]

    Mai, G., Janowicz, K., Hu, Y., Gao, S., Yan, B., Zhu, R., Cai, L., and Lao, N. (2022b). A review of location encoding for GeoAI : methods and applications. International Journal of Geographical Information Science , 36(4):639--673

    2022

  67. [75]

    Majic, I., Naghizade, E., Winter, S., and Tomko, M. (2021). RIM : a ray intersection model for the analysis of the between relationship of spatial objects in a 2D plane. International Journal of Geographical Information Science , 35(5):893--918

    2021

  68. [76]

    Majic, I., Wang, Z., Janowicz, K., and Karimi, M. (2024). Spatial task-explicity matters in prompting large multimodal models for spatial planning. In Proceedings of the 7th ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery , pages 99--105

    2024

  69. [77]

    Manvi, R., Khanna, S., Mai, G., Burke, M., Lobell, D., and Ermon, S. (2023). GeoLLM : Extracting geospatial knowledge from large language models. arXiv preprint arXiv:2310.06213

    2023 arXiv

  70. [78]

    Mark, D. M. and Egenhofer, M. J. (1994). Modeling spatial relations between lines and regions: combining formal mathematical models and human subjects testing. Cartography and geographic information systems , 21(4):195--212

    1994

  71. [79]

    Mooney, P., Cui, W., Guan, B., and Juh \'a sz, L. (2023). Towards understanding the geospatial skills of ChatGPT : Taking a geographic information systems ( GIS ) exam. In Proceedings of the 6th ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery , p...

    2023

  72. [80]

    Muennighoff, N. (2022). Sgpt: GPT sentence embeddings for semantic search . arXiv preprint arXiv:2202.08904

    2022 arXiv

  73. [81]

    M., Tworek, J., Yuan, Q., Tezak, N., Kim, J

    Neelakantan, A., Xu, T., Puri, R., Radford, A., Han, J. M., Tworek, J., Yuan, Q., Tezak, N., Kim, J. W., Hallacy, C., et al. (2022). Text and code embeddings by contrastive pre-training. arXiv preprint arXiv:2201.10005

    2022 arXiv

  74. [82]

    Introducing ChatGPT

    OpenAI (2022). Introducing ChatGPT

    2022

  75. [83]

    Radford, A., Narasimhan, K., Salimans, T., Sutskever, I., et al. (2018). Improving language understanding by generative pre-training. OpenAI blog , 0(1):12

    2018

  76. [84]

    Radford, A., Wu, J., Child, R., Luan, D., Amodei, D., Sutskever, I., et al. (2019). Language models are unsupervised multitask learners. OpenAI blog , 1(8):24

    2019

  77. [85]

    Ramesh, A., Pavlov, M., Goh, G., Gray, S., Voss, C., Radford, A., Chen, M., and Sutskever, I. (2021). DALL·E : Creating images from text

    2021

  78. [86]

    A., Cui, Z., and Cohn, A

    Randell, D. A., Cui, Z., and Cohn, A. G. (1992). A spatial logic based on regions and connection. Principles of Knowledge Representation and Reasoning , 92:165--176

    1992

  79. [87]

    Rao, J., Gao, S., Mai, G., and Janowicz, K. (2023). Building privacy-preserving and secure geospatial artificial intelligence foundation models (vision paper). In Proceedings of the 31st ACM International Conference on Advances in Geographic Information Systems , pages 1--4

    2023

  80. [88]

    and Gurevych, I

    Reimers, N. and Gurevych, I. (2019). Sentence-BERT : Sentence embeddings using siamese BERT-networks . arXiv preprint arXiv:1908.10084

    2019 arXiv

  81. [89]

    M., Egenhofer, M

    Reis, R. M., Egenhofer, M. J., and Matos, J. L. (2008). Conceptual neighborhoods of topological relations between lines. In Headway in Spatial Data Handling: 13th International Symposium on Spatial Data Handling , pages 557--574. Springer

    2008

  82. [90]

    and Nebel, B

    Renz, J. and Nebel, B. (1998). Spatial reasoning with topological information. In Spatial Cognition: An Interdisciplinary Approach to Representing and Processing Spatial Knowledge , pages 351--371. Springer

    1998

  83. [91]

    and Urrutia, J

    Sack, J.-R. and Urrutia, J. (1999). Handbook of computational geometry . Elsevier

    1999

  84. [92]

    Scheider, S., Nyamsuren, E., Kruiger, H., and Xu, H. (2021). Geo-analytical question-answering with GIS . International Journal of Digital Earth , 14(1):1--14

    2021

  85. [93]

    Shannon, C. E. (1948). A mathematical theory of communication. The Bell system technical journal , 27(3):379--423

    1948

  86. [94]

    Sheth, A., Roy, K., and Gaur, M. (2023). Neurosymbolic artificial intelligence (why, what, and how). IEEE Intelligent Systems , 38(3):56--62

    2023

  87. [95]

    Skoumas, G., Pfoser, D., Kyrillidis, A., and Sellis, T. (2016). Location estimation using crowdsourced spatial relations. ACM Transactions on Spatial Algorithms and Systems (TSAS) , 2(2):1--23

    2016

  88. [96]

    and Xu, J

    Tao, R. and Xu, J. (2023). Mapping with ChatGPT . ISPRS International Journal of Geo-Information , 12(7):284

    2023

  89. [97]

    Tellex, S., Kollar, T., Dickerson, S., Walter, M., Banerjee, A., Teller, S., and Roy, N. (2011). Understanding natural language commands for robotic navigation and mobile manipulation. In Proceedings of the AAAI Conference on Artificial Intelligence , volume 25, pages 1507--1514

    2011

  90. [98]

    Thoppilan, R., De Freitas, D., Hall, J., Shazeer, N., Kulshreshtha, A., Cheng, H.-T., Jin, A., Bos, T., Baker, L., Du, Y., et al. (2022). LaMDA : Language models for dialog applications. arXiv preprint arXiv:2201.08239

    2022 arXiv

  91. [99]

    Tucker, S. (2024). A systematic review of geospatial location embedding approaches in large language models: A path to spatial ai systems. arXiv preprint arXiv:2401.10279

    2024 arXiv

  92. [100]

    O., Klippel, A., and Baldwin, T

    Wallgr \"u n, J. O., Klippel, A., and Baldwin, T. (2014). Building a corpus of spatial relational expressions extracted from web documents. In Proceedings of the 8th workshop on geographic information retrieval , pages 1--8

    2014

  93. [101]

    O., Klippel, A., and Karimzadeh, M

    Wallgr \"u n, J. O., Klippel, A., and Karimzadeh, M. (2015). Towards contextualized models of spatial relations. In Proceedings of the 9th Workshop on Geographic Information Retrieval , pages 1--2

    2015

  94. [102]

    Wang, F. (2000). A fuzzy grammar and possibility theory-based natural language user interface for spatial queries. Fuzzy sets and systems , 113(1):147--159

    2000

  95. [103]

    Wang, L., Yang, N., and Wei, F. (2023). Query2doc: Query expansion with large language models. arXiv preprint arXiv:2303.07678

    2023 arXiv

  96. [104]

    Wang, Z., Ye, X., and Tsou, M.-H. (2016). Spatial, temporal, and content analysis of Twitter for wildfire hazards. Natural Hazards , 83:523--540

    2016

  97. [105]

    V., Zhou, D., et al

    Wei, J., Wang, X., Schuurmans, D., Bosma, M., Xia, F., Chi, E., Le, Q. V., Zhou, D., et al. (2022). Chain-of-thought prompting elicits reasoning in large language models. Advances in Neural Information Processing Systems , 35:24824--24837

    2022

  98. [106]

    and Wallgr \"u n, J

    Wolter, D. and Wallgr \"u n, J. O. (2012). Qualitative spatial reasoning for applications: New challenges and the sparq toolbox. Qualitative Spatio-Temporal Representation and Reasoning: Trends and Future Directions: Trends and Future Directions , page 336

    2012

  99. [107]

    Wu, K., Zhang, X., Dang, Y., and Ye, P. (2023a). Deep learning models for spatial relation extraction in text. Geo-spatial Information Science , 26(1):58--70

    2023

  100. [108]

    Wu, M., Huang, Q., Gao, S., and Zhang, Z. (2023b). Mixed land use measurement and mapping with street view images and spatial context-aware prompts via zero-shot multimodal learning. International Journal of Applied Earth Observation and Geoinformation , 125:103591

    2023

  101. [109]

    Xie, Y., Wang, Z., Mai, G., Li, Y., Jia, X., Gao, S., and Wang, S. (2023). Geo-foundation models: Reality, gaps and opportunities. In Proceedings of the 31st ACM International Conference on Advances in Geographic Information Systems , pages 1--4

    2023

  102. [110]

    K., Kasai, J., and Yildirim, I

    Yamada, Y., Bao, Y., Lampinen, A. K., Kasai, J., and Yildirim, I. (2023). Evaluating spatial understanding of large language models. arXiv preprint arXiv:2310.14540

    2023 arXiv

  103. [111]

    Yan, B., Janowicz, K., Mai, G., and Gao, S. (2017). From ITDL to Place2Vec : Reasoning about place type similarity and relatedness by learning embeddings from augmented spatial contexts. In Proceedings of the 25th ACM SIGSPATIAL international conference on advances in geograph...

    2017

  104. [112]

    W., Han, R., Fei-Fei, L., and Xie, S

    Yang, J., Yang, S., Gupta, A. W., Han, R., Fei-Fei, L., and Xie, S. (2024). Thinking in space: How multimodal large language models see, remember, and recall spaces. arXiv preprint arXiv:2412.14171

    2024 arXiv

  105. [113]

    C., Smith, K

    Yang, X., Chen, A., PourNejatian, N., Shin, H. C., Smith, K. E., Parisien, C., Compas, C., Martin, C., Costa, A. B., Flores, M. G., et al. (2022). A large language model for electronic health records. npj Digital Medicine , 5(1):194

    2022

  106. [114]

    Yuan, Y. (2011). Extracting spatial relations from document for geographic information retrieval. In 2011 19th International Conference on Geoinformatics , pages 1--5. IEEE

    2011

  107. [115]

    Zhang, Q., Kang, Y., and Roth, R. (2023). The Ethics of AI-Generated Maps: DALLE-2 and AI’s Implications for Cartography (Short Paper) . In 12th International conference on geographic information science (GIScience 2023) , pages 93--98. Schloss Dagstuhl--Leibniz-Zentrum f \"u ...

    2023

  108. [116]

    Zhang, Y., Wei, C., He, Z., and Yu, W. (2024). GeoGPT: An assistant for understanding and processing geospatial tasks . International Journal of Applied Earth Observation and Geoinformation , 131:103976

    2024

  109. [117]

    Zheng, O., Abdel-Aty, M., Wang, D., Wang, Z., and Ding, S. (2023). ChatGPT is on the horizon: Could a large language model be all we need for intelligent transportation? arXiv preprint arXiv:2303.05382

    2023 arXiv

  110. [118]

    Zhu, R., Janowicz, K., Cai, L., and Mai, G. (2022). Reasoning over higher-order qualitative spatial relations via spatially explicit neural networks. International Journal of Geographical Information Science , 36(11):2194--2225

    2022

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.