Pith. sign in

REVIEW 4 major objections 5 minor 50 references

GeoVLM: Improving Automated Vehicle Geolocalisation Using Vision-Language Matching

T0 review · 4 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 GeoVLM improves top-1 accuracy in cross-view geo-localisation by fusing language descriptions with image embeddings in a trainable reranking stage, raising R@1 on VIGOR same-area from 77.86% to 85.64% and on cross-area from 61.70% to…

desk verdict Clean idea with real R@1 gains on VIGOR, but the paper never isolates whether language or just the trainable reranker is responsible. read the letter →

arxiv 2505.13669 v1 pith:OEC5L33X submitted 2025-05-19 cs.CV cs.RO

Barkin Dagda , Muhammad Awais , Saber Fallah This is my paper · ORCID
classification cs.CVcs.RO
keywords cross-viewgeo-localisationvision-languagemodelsrerankingimageretrievalautomatedvehiclelocalisationzero-shotcaptioningUKdataset
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

Cross-view geo-localisation matches a street-level image to the correct satellite tile, and existing systems often find the right tile within their top ten but fail to rank it first. This paper proposes GeoVLM, a trainable reranking stage that uses zero-shot vision-language descriptions of each scene, generated by answering thirty multiple-choice questions, combined with the usual image embeddings to reorder the top-10 candidates. The central claim is that the language channel disambiguates visually similar locations, and the reported results support it: top-1 accuracy rises from 77.86% to 85.64% on VIGOR same-area and from 61.70% to 71.98% cross-area, with smaller gains on University-1652 and on a new UK driving dataset called CVUK. A sympathetic reader would care because top-1 accuracy is what matters when a vehicle uses the retrieval result directly as its position estimate.

What carries the argument

The load-bearing object is the cross-embedding reranking module: a small trainable network that adds a query's projected image and text embeddings to each candidate's projected image and text embeddings, runs the sum through a cross-embedding aligner, and applies a sigmoid to produce a reranking score. The text side is created by a fixed set of 30 multiple-choice questions answered by BLIP-2, with answers inserted into a sentence template to form structured captions; these captions are embedded with OpenAI's text-embedding-3-small model. The module is trained with a margin-based ranking loss (margin m=1) that only activates when a negative candidate's score is too close to or above the positive score, so training focuses on reordering the already-retrieved top-10 instead of learning visual features from scratch.

What would settle it

Retrain the reranking module under the exact same protocol but replace each caption with a random permutation of its own words (same vocabulary, different order); if R@1 on VIGOR same-area stays at 85.64% instead of falling back toward the 77.86% Sample4Geo baseline, the semantic content of the language descriptions is not what drives the improvement.

Watch

Extended reading notes

Core claim

GeoVLM splits the task into retrieval and reranking. A frozen pretrained cross-view encoder, Sample4Geo, retrieves the ten most similar satellite images for each query by cosine similarity. A trainable reranking module then scores each query-candidate pair: the query and candidate image embeddings and their caption embeddings are linearly projected into a shared space, summed element-wise, passed through fully connected layers with layer normalization and ReLU, and finally through a sigmoid to yield a match score. The captions are produced offline by BLIP-2 answering 30 multiple-choice questions about environment, roads, architecture, and colors, assembled into a structured template to keep them consistent. The paper's discovery is that this fusion raises exact top-1 match accuracy on all three benchmarks, and the authors attribute the gain to language descriptions that preserve scene attributes across viewpoints and help tell apart visually similar places.

Load-bearing premise

The reranking signal depends on the captions being stable and carrying the same scene attributes across ground and satellite views; the paper's own stability check reports only 0.44 Jaccard word overlap between repeated descriptions, so if the captions are largely noisy or viewpoint-specific, the learned reranker would have little reliable signal to exploit.

Editorial extensions

If this is right

  • On VIGOR same-area, R@1 improves from 77.86% to 85.64% while R@10 stays fixed at 97.21%, showing the reranking reorders the existing shortlist rather than finding new candidates.
  • On VIGOR cross-area, R@1 improves from 61.70% to 71.98%, indicating the language-guided reranking transfers to cities not seen during training.
  • On the new CVUK dataset at the 0.5 km positional-error threshold, R@1 rises from 11.23% to 13.50%, and on University-1652 Drone2Sat it rises from 92.65% to 93.15%.
  • Since caption generation takes 8.3 seconds per image and is done offline, the approach is not real-time; any on-vehicle deployment would need to move description generation off the critical path or adopt a faster VLM.

Reading between the lines

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

  • The same two-stage recipe—high-recall retrieval plus a language-aware reranker—could be applied to other retrieval tasks whose bottleneck is top-1 rather than top-10, such as landmark recognition or drone re-identification, by swapping the base encoder and the caption generator.
  • Given that repeated captions share only 0.44 Jaccard word overlap, ensembling several independently generated captions per image before embedding might make the text signal more stable and could push R@1 higher than the single-caption version.
  • The paper's observation that caption embeddings cluster by city (Figure 4) suggests a pure text classifier could serve as a hard geographical prior to prune candidates, which would be especially useful in dense urban areas.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes GeoVLM, a two-stage cross-view geo-localisation method. In the first phase, a frozen Sample4Geo encoder retrieves the top-10 satellite candidates for each ground query. In the second phase, a trainable reranking module fuses image embeddings with text embeddings of captions that are pre-generated by BLIP-2 answering 30 multiple-choice questions about each ground and satellite image. The reranking module is trained with a margin-based ranking loss and produces a score used to reorder the top-10 candidates. The method is evaluated on VIGOR, University-1652, and a new UK-based dataset called CVUK. The authors report improved R@1 on VIGOR same-area and cross-area settings, improved R@1 on University-1652, and improved positional accuracy on CVUK, and they claim these are the first explainable cross-view geo-localisation method and the first UK cross-view dataset.

Significance. If the central claim is supported, the paper opens a useful direction: using zero-shot vision-language descriptions as a complementary signal for disambiguating visually similar cross-view scenes. The idea is well motivated by the known weakness of existing methods in top-1 accuracy, and the paper provides a concrete, reproducible pipeline with code released. The introduction of the CVUK dataset, including seasonal and lighting variation, is a positive contribution. The systematic MCQ-based caption generation is also a strength. However, the evidence as presented does not yet isolate the contribution of the language modality from the contribution of the trainable reranking head, and at least one reported result contradicts the paper's stated claims. The significance of the work therefore depends on additional controlled experiments.

major comments (4)
  1. [Section III-C, Eq. (3); Tables II-IV] The central claim that language descriptions improve reranking is not supported by the reported experiments because no ablation isolates the text modality. GeoVLM differs from the Sample4Geo baseline in two ways simultaneously: it adds a trainable reranking head with a learned projection and scoring function, and it adds text embeddings to that head. Comparing GeoVLM to Sample4Geo therefore conflates these two changes. To establish that the VLM-generated descriptions are responsible for the improvement, the authors should train the same reranking architecture with (i) image embeddings only, (ii) text embeddings only, and (iii) image plus text, and report all three on the same benchmarks. Without such a text-modality ablation, the reported gains could come entirely from the extra learned parameters exploiting the base encoder's top-10 structure rather than from language.
  2. [Abstract, Contributions, and Table II] The paper claims in the abstract and in the contribution list that GeoVLM improves top-1 and top-5 accuracy on VIGOR, but Table II shows that in the cross-area setting R@5 drops from 83.50 (Sample4Geo) to 81.87 (GeoVLM), while R@10 is unchanged at 88.00. This is an internal inconsistency in a headline claim. The authors should either qualify the claim to top-1 (and same-area top-5) or explain why a decrease in cross-area top-5 is consistent with the proposed method's objective. This is not a presentation nit; it affects how the contribution is stated.
  3. [Tables II-IV and Section IV-C] No error bars, repeated runs, or statistical significance tests are reported. Several of the claimed gains are small, notably University-1652 Drone2Sat R@1 (92.65 to 93.15, a 0.5 percentage-point difference) and Sat2Drone AP (91.39 to 93.36). With a single run and no variance estimate, these differences could be within run-to-run noise. The authors should report means and standard deviations over at least three random seeds, or otherwise justify that the differences are stable.
  4. [Section IV-D, Table III] The CVUK evaluation protocol is under-specified, making the results difficult to interpret. The text says CVUK is used for 'fine-tuning and validating the pre-trained models', and Table III reports both VIGOR-to-CVUK and CVUK-to-VIGOR numbers, but the paper does not specify how the CVUK data are split into training and test sets, whether the GeoVLM reranking module itself is trained or only the backbone is fine-tuned on CVUK, or how the 0.5 km positional-error threshold interacts with the statement that satellite images are at least 100 meters apart. These details are needed to assess whether the CVUK results are a fair evaluation or a fine-tuning demonstration.
minor comments (5)
  1. [Section IV-G, Table V] The stability check reports Jaccard similarity of 0.44 between two generations of the same image's description, which is described as 'moderate overlap'. Since captions are generated once and then fixed, high word-level variability across repeated generations may not directly affect inference, but it does raise a question about whether the MCQ set actually constrains the model as claimed. A brief discussion of why Jaccard 0.44 is acceptable for the method would be useful.
  2. [Section IV-F, Figure 4] Figure 4 shows t-SNE-like clusters of description embeddings, but the axes and embedding model are not labeled. Please state the projection method and the source of the embeddings, and clarify what the reader should conclude from the clustering.
  3. [Section III-C, Eq. (3)] Equation (3) uses notation I'_q + C'_q as a sum of projected image and text embeddings, but the dimensions are not stated to be equal after projection, nor is the projection dimensionality given. Please specify the projected dimension and confirm that the element-wise addition is performed after aligning dimensions.
  4. [Section IV-C] Training details for the reranking module are missing: optimizer, learning rate, batch size, number of epochs, and how the positive/negative pairs are sampled from the top-10 candidates are not reported. These details are necessary for reproducibility.
  5. [Throughout] There are several typographical issues, e.g., 'canditates' in Section III-C, 'SampleGeo' instead of 'Sample4Geo' in Section IV-D, and inconsistent use of 'geo-localisation' vs 'geolocalization'. A careful proofread is needed.

Circularity Check

0 steps flagged · score 1.0 of 10

No material circularity: GeoVLM's reported gains come from training a reranking module on standard benchmarks and testing on held-out splits, not from a derivation that reduces to its own inputs.

full rationale

The central claim is that a trainable reranking module, fusing Sample4Geo image embeddings with BLIP-2-generated captions, improves top-match accuracy over a frozen cross-view encoder. The derivation chain is empirical in both phases: Phase 1 uses a frozen external backbone (Sample4Geo) to retrieve top-10 candidates, and Phase 2 trains the parameters W and b in Eq. 3 using the margin-based ranking loss in Eq. 4. Results are then reported on held-out evaluation splits of VIGOR, University-1652, and the newly introduced CVUK set. No equation or protocol in the paper defines the reported improvement as a restatement of the fitted parameters, and the caption-generation pipeline is a fixed, external process using BLIP-2 and OpenAI text embeddings. The only self-citation, HighwayLLM [29], appears in related work as background motivation and is not load-bearing for the GeoVLM result. The paper's known weakness is the absence of an image-only reranking ablation, so the specific contribution of the text modality is not experimentally isolated; however, a missing controlled comparison is an experimental-design issue, not circularity. The empirical improvements are self-contained against external benchmarks, so the appropriate circularity score is low.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the quality of auto-generated captions and on the base encoder's ability to include the true match in the top ten. Neither is shown to be stable or guaranteed, and no fitted constants define the reported metrics.

free parameters (3)
  • Margin m in ranking loss = 1
    Set by hand in Eq. 4; controls how much the positive score must exceed negative scores. A different margin would change the training signal and final rankings.
  • Number of candidates k = 10
    Reranking is applied to the top-10 retrieved by the base encoder, matching R@10 of Sample4Geo. This design choice bounds the maximum achievable recall.
  • Number of MCQs = 30
    The question set size is chosen by the authors and affects caption content and stability; no ablation is provided.
assumptions (4)
  • domain assumption BLIP-2 VQA answers to the 30 MCQs are accurate and stable for both ground and satellite images.
    The reranking signal is built entirely from these captions. The stability check reports cosine similarity 0.83 and Jaccard similarity 0.44, so answers vary significantly across runs. See Section III-B and Table V.
  • ad hoc to paper The MCQ set and sentence template, designed with ChatGPT, capture scene attributes that are invariant and discriminative across viewpoints.
    There is no systematic validation that the selected attributes (road layout, roof colors, vegetation) transfer reliably from street level to satellite view. This is introduced specifically for this paper in Section III-B.
  • ad hoc to paper OpenAI text-embedding-3-small embeddings, after linear projection, can be additively fused with image embeddings in Eq. 3.
    The element-wise addition of projected text and image features assumes the two modalities are comparable in the shared space, a modeling choice not independently justified.
  • domain assumption The pretrained Sample4Geo encoder retrieves the correct match in the top-10 for nearly all test queries.
    If the correct match is absent from the top-10, GeoVLM cannot recover it. This is acknowledged in the paper; VIGOR R@10 is 97.21 same-area but only 88.00 cross-area, so the method fails on 12% of cross-area queries by construction.

how reviews work

0 comments
Cite this review

Pith. "Pith review of GeoVLM: Improving Automated Vehicle Geolocalisation Using Vision-Language Matching." pith.science (2026). https://pith.science/paper/OEC5L33X

@misc{pith2026250513669,
  author       = {Pith},
  title        = {Pith review of: GeoVLM: Improving Automated Vehicle Geolocalisation Using Vision-Language Matching},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OEC5L33X}},
  note         = {Machine review of arXiv:2505.13669}
}
read the original abstract

Cross-view geo-localisation identifies coarse geographical position of an automated vehicle by matching a ground-level image to a geo-tagged satellite image from a database. Despite the advancements in Cross-view geo-localisation, significant challenges still persist such as similar looking scenes which makes it challenging to find the correct match as the top match. Existing approaches reach high recall rates but they still fail to rank the correct image as the top match. To address this challenge, this paper proposes GeoVLM, a novel approach which uses the zero-shot capabilities of vision language models to enable cross-view geo-localisation using interpretable cross-view language descriptions. GeoVLM is a trainable reranking approach which improves the best match accuracy of cross-view geo-localisation. GeoVLM is evaluated on standard benchmark VIGOR and University-1652 and also through real-life driving environments using Cross-View United Kingdom, a new benchmark dataset introduced in this paper. The results of the paper show that GeoVLM improves retrieval performance of cross-view geo-localisation compared to the state-of-the-art methods with the help of explainable natural language descriptions. The code is available at https://github.com/CAV-Research-Lab/GeoVLM

Figures

Figures reproduced from arXiv: 2505.13669 by the authors.

Figure 1
Figure 1. Overview of proposed GeoVLM pipeline for CVGL. The left side of the figure (blue section) illustrates Phase 1, where the network is frozen and top-k [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 3
Figure 3. Visualization of a correct match from the CVUK dataset with [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Visual clusters of image descriptions for various locations in the [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figures from the paper (1 more)
Figure 2
Figure 2. Figure 2: Visualization of cross-view images with their corresponding captions [PITH_FULL_IMAGE:figures/full_fig_p007_2.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 25 canonical work pages

  1. [1]

    Autonomous vehicles navigation with visual target tracking: Technical approaches.Algorithms, 1:153–182, 2008

    Zhen Jia, Arjuna Balasuriya, and Subhash Challa. Autonomous vehicles navigation with visual target tracking: Technical approaches.Algorithms, 1:153–182, 2008. https://api.semanticscholar.org/CorpusID:1735298

    work page 2008

  2. [2]

    Spatial-aware feature aggregation for cross-view image based geo-localisation,

    Y . Shi, L. Liu, X. Yu, and H. Li, “Spatial-aware feature aggregation for cross-view image based geo-localisation,” in Proc. 33rd Conf. Neural Inf. Process. Syst. (NeurIPS) , Red Hook, NY , USA: Curran Associates Inc., 2019, pp. 2, 6, 7

    work page 2019

  3. [3]

    S. Zhu, M. Shah, and C. Chen, ”TransGeo: Transformer Is All You Need for Cross-view Image Geo-localization,” in 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , Los Alamitos, CA, USA: IEEE Computer Society, Jun. 2022, pp. 1152–1161. doi: 10.1109/CVPR52688.2022.00123

    arXiv 2022

  4. [4]

    Fervers, S

    F. Fervers, S. Bullinger, C. Bodensteiner, M. Arens, and R. Stiefelhagen, ”Uncertainty-Aware Vision-Based Metric Cross-view Geolocalization,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2023, pp. 21621-21631

    work page 2023

  5. [5]

    Zhang, X

    X. Zhang, X. Li, W. Sultani, Y . Zhou, and S. Wshah, ”Cross-view geo- localization via learning disentangled geometric layout correspondence,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 37, no. 3, 2023, pp. 25457. doi: 10.1609/aaai.v37i3.25457

    doi:10.1609/aaai.v37i3.25457 2023

  6. [6]

    Wang, X., Xu, R., Cui, Z., Wan, Z., and Zhang, Y . (2023). Fine-Grained Cross-View Geo-Localization Using a Correlation-Aware Homography Estimator

    work page 2023

  7. [7]

    Gao, H., Wang, Z., Li, Y ., Long, K., Yang, M., and Shen, Y . (2024). A Survey for Foundation Models in Autonomous Driving

    work page 2024

  8. [8]

    [Online]

    OpenAI, ”ChatGPT,” 2023. [Online]. Available: https://openai.com/blog/chatgpt/

    work page 2023

Show all 50 references
  1. [9]

    Touvron, T

    H. Touvron, T. Lavril, G. Izacard, X. Martinet, M.-A. Lachaux, T. Lacroix, B. Rozi `ere, N. Goyal, E. Hambro, F. Azhar, A. Rodriguez, A. Joulin, E. Grave, and G. Lample, ”LLaMA: Open and Efficient Foundation Language Models,” ArXiv, vol. abs/2302.13971, 2023

    2023 arXiv

  2. [10]

    Touvron, L

    H. Touvron, L. Martin, K. Stone, P. Albert, A. Almahairi, Y . Babaei, N. Bashlykov, S. Batra, P. Bhargava, S. Bhosale, et al., ”Llama 2: Open Foundation and Fine-Tuned Chat Models,” arXiv:2307.09288, 2023b

    arXiv

  3. [11]

    Xu et al., ”DriveGPT4: Interpretable End-to-End Autonomous Driving Via Large Language Model,” IEEE Robotics and Au- tomation Letters, vol

    Z. Xu et al., ”DriveGPT4: Interpretable End-to-End Autonomous Driving Via Large Language Model,” IEEE Robotics and Au- tomation Letters, vol. 9, no. 10, pp. 8186-8193, Oct. 2024, doi: 10.1109/LRA.2024.3440097. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 9

    2024

  4. [12]

    Lewis, E

    P. Lewis, E. Perez, A. Piktus, F. Petroni, V . Karpukhin, N. Goyal, H. K¨uttler, M. Lewis, W.-t. Yih, T. Rockt ¨aschel, S. Riedel, and D. Kiela, ”Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks,” in Advances in Neural Information Processing Systems (NeurIPS), ...

    2020

  5. [13]

    A. Q. Jiang, A. Sablayrolles, A. Roux, A. Mensch, B. Savary, C. Bamford, D. S. Chaplot, D. de las Casas, E. Bou Hanna, F. Bressand, G. Lengyel, G. Bour, G. Lample, L. Renard Lavaud, L. Saulnier, M.-A. Lachaux, P. Stock, S. Subramanian, S. Yang, S. Antoniak, T. Le Scao, T. Gerv...

    2024 arXiv

  6. [14]

    Radford, J

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark, G. Krueger, and I. Sutskever, ”Learning Transferable Visual Models From Natural Language Supervi- sion,” in International Conference on Machine Learning (ICML) , 2021

    2021

  7. [15]

    Y . Li, F. Liang, L. Zhao, Y . Cui, W. Ouyang, J. Shao, F. Yu, and J. Yan, ”Supervision Exists Everywhere: A Data Efficient Contrastive Language-Image Pre-training Paradigm,” in International Conference on Learning Representations (ICLR) , 2022

    2022

  8. [16]

    J. Li, D. Li, C. Xiong, and S. Hoi, ”BLIP: Bootstrapping Language- Image Pre-training for Unified Vision-Language Understanding and Generation,” arXiv:2201.12086, 2022

    2022 arXiv

  9. [17]

    J. Li, D. Li, S. Savarese, and S. Hoi, ”BLIP-2: Bootstrapping Language- Image Pre-training with Frozen Image Encoders and Large Language Models,” arXiv preprint arXiv:2301.12597 , 2023

    2023 arXiv

  10. [18]

    Brohan et al., ”RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control,” arXiv:2307.15818, 2023

    A. Brohan et al., ”RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control,” arXiv:2307.15818, 2023

    2023 arXiv

  11. [19]

    Wen et al., ”On the Road with GPT-4V(ision): Early Explorations of Visual-Language Model on Autonomous Driving,” arXiv:2311.05332, 2023

    L. Wen et al., ”On the Road with GPT-4V(ision): Early Explorations of Visual-Language Model on Autonomous Driving,” arXiv:2311.05332, 2023

    2023 arXiv

  12. [20]

    Shi and H

    Y . Shi and H. Li, ”Beyond Cross-view Image Retrieval: Highly Accurate Vehicle Localization Using Satellite Image,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 8445–8454

    2022

  13. [21]

    SMDT: Cross- View Geo-Localization with Image Alignment and Transformer,

    X. Tian, J. Shao, D. Ouyang, A. Zhu, and F. Chen, “SMDT: Cross- View Geo-Localization with Image Alignment and Transformer,” in 2022 IEEE International Conference on Multimedia and Expo (ICME) , 2022, pp. 1–6. DOI: 10.1109/ICME52920.2022.9859992

    2022

  14. [22]

    Guided Curriculum Model Adaptation and Uncertainty-Aware Evaluation for Semantic Nighttime Image Segmentation,

    C. Sakaridis, D. Dai, and L. Van Gool, “Guided Curriculum Model Adaptation and Uncertainty-Aware Evaluation for Semantic Nighttime Image Segmentation,” in Proceedings of the IEEE International Confer- ence on Computer Vision (ICCV) , 2019

    2019

  15. [23]

    S. Xiao, Z. Liu, P. Zhang, N. Muennighoff, D. Lian, and J.-Y . Nie, ”C- Pack: Packaged Resources To Advance General Chinese Embedding,” in Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval , 2024, pp. 1234– 1243

    2024

  16. [24]

    Y . Zhu, H. Yang, Y . Lu, and Q. Huang, ”Simple, Effective and Gen- eral: A New Backbone for Cross-view Image Geo-localization,” arXiv preprint arXiv:2302.01572, 2023

    2023 arXiv

  17. [25]

    How Good Are Low-bit Quantized LLaMA3 Models? An Empirical Study,

    W. Huang, X. Ma, H. Qin, X. Zheng, C. Lv, H. Chen, J. Luo, X. Qi, X. Liu, and M. Magno, “How Good Are Low-bit Quantized LLaMA3 Models? An Empirical Study,” 2024,

    2024

  18. [26]

    Deuser, K

    F. Deuser, K. Habel, and N. Oswald, ”Sample4Geo: Hard Negative Sampling For Cross-View Geo-Localisation,” in Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , 2023, pp. 1146–1155

    2023

  19. [27]

    Y . Jin, R. Yang, Z. Yi, X. Shen, H. Peng, X. Liu, J. Qin, J. Li, J. Xie, P. Gao, G. Zhou, and J. Gong, ”SurrealDriver: Designing LLM- powered Generative Driver Agent Framework based on Human Drivers’ Driving-thinking Data,” in 2024 IEEE/RSJ International Conference on Intelli...

    2024

  20. [28]

    OpenAI, Achiam, J., Adler, S., et al. (2024). GPT-4 Technical Report. arXiv preprint arXiv:2303.08774 . Retrieved from https://arxiv.org/abs/ 2303.08774

    2024 arXiv

  21. [29]

    Yildirim, M., Dagda, B., and Fallah, S. (2024). HighwayLLM: Decision- Making and Navigation in Highway Driving with RL-Informed Lan- guage Model. arXiv preprint arXiv:2405.13547 . Retrieved from https: //arxiv.org/abs/2405.13547

    2024 arXiv

  22. [30]

    Zheng, L

    Z. Zheng, L. Shen, L. Tian, Z. Wang, Y . Wang, Y . Tian, and S. Wang, ”University-1652: A Multi-view Multi-source Benchmark for Drone- based Geo-localization,” ACM Transactions on Multimedia Computing, Communications, and Applications (TOMM) , vol. 16, no. 3, pp. 1–20, 2020

    2020

  23. [31]

    Y . Hu, Y . Liu, and B. Hui, ”Combining OpenStreetMap with Satellite Imagery to Enhance Cross-View Geo-Localization,” Sensors (Basel, Switzerland), vol. 25, 2024

    2024

  24. [32]

    J. Ye, Z. Lv, W. Li, J. Yu, H. Yang, H. Zhong, and C. He, ”Cross-view image geo-localization with Panorama-BEV Co-Retrieval Network,” in Proceedings of the European Conference on Computer Vision (ECCV) , 2024

    2024

  25. [33]

    Ghanem, A

    A. Ghanem, A. S. Abdelhay, N. E. Salah, A. Y . N. Eldeen, M. Elhenawy, M. Masoud, A. M. Hassan, and A. A. Hassan, ”Leveraging cross-view geo-localization with ensemble learning and temporal awareness,” PLOS ONE, vol. 18, 2023

    2023

  26. [34]

    Chen, Z.-X

    Z. Chen, Z.-X. Yang, and H.-J. Rong, ”Without Paired Labeled Data: An End-to-End Self-Supervised Paradigm for UA V-View Geo-Localization,” in Proceedings of an International Conference , 2025

    2025

  27. [35]

    H. Li, C. Xu, W. Yang, H. Yu, and G.-S. Xia, ”Learning Cross-View Visual Geo-Localization Without Ground Truth,” IEEE Transactions on Geoscience and Remote Sensing , vol. 62, pp. 1-17, 2024

    2024

  28. [36]

    Workman, R

    S. Workman, R. Souvenir, and N. Jacobs, ”Wide-Area Image Geolo- calization with Aerial Reference Imagery,” in Proceedings of the IEEE International Conference on Computer Vision (ICCV) , 2015, pp. 3961- 3969

    2015

  29. [37]

    Choudhary, V

    T. Choudhary, V . Dewangan, S. Chandhok, S. Priyadarshan, A. Jain, A. K. Singh, S. Srivastava, K. M. Jatavallabhula, and K. M. Krishna, ”Talk2BEV: Language-enhanced Bird’s-eye View Maps for Autonomous Driving,” in 2024 IEEE International Conference on Robotics and Automation (...

    2024

  30. [38]

    S. Park, M. Lee, J. Kang, H. Choi, Y . Park, J. Cho, A. Lee, and D. Kim, ”VLAAD: Vision and Language Assistant for Autonomous Driv- ing,” in *Proc. IEEE/CVF Winter Conf. Appl. Comput. Vis. (W ACV) Workshops*, Jan. 2024, pp. 980-987

    2024

  31. [39]

    J. Yuan, S. Sun, D. Omeiza, B. Zhao, P. Newman, L. Kunze, and M. Gadd, ”RAG-Driver: Generalisable Driving Explanations with Retrieval-Augmented In-Context Learning in Multi-Modal Large Lan- guage Model,” arXiv preprint arXiv:2402.10828, 2024

    2024

  32. [40]

    M. Peng, X. Guo, X. Chen, K. Chen, and M. Zhu, ”LC-LLM: Ex- plainable lane-change intention and trajectory predictions with Large Language Models,” Communications in Transportation Research , vol. 5, pp. 100170, 2025. doi: 10.1016/j.commtr.2025.100170

    2025

  33. [41]

    C. Sima, K. Renz, K. Chitta, L. Chen, H. Zhang, C. Xie, J. Beißwenger, P. Luo, A. Geiger, and H. Li, ”DriveLM: Driving with Graph Visual Question Answering,” arXiv preprint arXiv:2312.14150, 2023. [Online]. Available: https://arxiv.org/abs/2312.14150

    2023 arXiv

  34. [42]

    Y . Cui, S. Huang, J. Zhong, Z. Liu, Y . Wang, C. Sun, B. Li, X. Wang, and A. Khajepour, ”DriveLLM: Charting the Path Toward Full Autonomous Driving With Large Language Models,” IEEE Transac- tions on Intelligent Vehicles, vol. 9, no. 1, pp. 1450-1464, 2024, doi: 10.1109/TIV ....

    2024

  35. [44]

    H. Sha, Y . Mu, Y . Jiang, L. Chen, C. Xu, P. Luo, S. E. Li, M. Tomizuka, W. Zhan, and M. Ding, ”LanguageMPC: Large Language Models as Decision Makers for Autonomous Driving,” arXiv preprint arXiv:2310.03026, 2023. [Online]. Available: https://arxiv.org/abs/2310. 03026

    2023 arXiv

  36. [45]

    K. Long, H. Shi, J. Liu, and X. Li, ”VLM-MPC: Vision Language Foun- dation Model (VLM)-Guided Model Predictive Controller (MPC) for Autonomous Driving,” arXiv preprint arXiv:2408.04821, 2024. [Online]. Available: https://arxiv.org/abs/2408.04821

    2024 arXiv

  37. [46]

    C. Pan, B. Yaman, T. Nesti, A. Mallik, A. G. Allievi, S. Velipasalar, and L. Ren, ”VLP: Vision Language Planning for Autonomous Driving,” arXiv preprint arXiv:2401.05577, 2024. [Online]. Available: https:// arxiv.org/abs/2401.05577

    2024 arXiv

  38. [47]

    Deuser, K

    F. Deuser, K. Habel, M. Werner, and N. Oswald, ”Orientation-Guided Contrastive Learning for UA V-View Geo-Localisation,” in UA VM ’23, 2023, pp. 7-11

    2023

  39. [48]

    T. Wang, Z. Zheng, Z. Zhu, Y . Gao, Y . Yang, and C. Yan, ”Learning Cross-view Geo-localization Embeddings via Dynamic Weighted Decor- relation Regularization,” arXiv:2211.05296, 2022

    2022 arXiv

  40. [49]

    H. Yang, X. Lu, and Y . Zhu, ”Cross-view Geo-localization with Evolving Transformer,” arXiv:2107.00842, 2021

    2021 arXiv

  41. [50]

    Nayyeri, M., Zhou, X., Vahdati, S., Shariat Yazdi, H., & Lehmann, J. (2019). Adaptive Margin Ranking Loss for Knowledge Graph Embeddings via a Correntropy Objective Function. arXiv preprint arXiv:1907.05336. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 10 A. A PPE...

    2019 arXiv

  42. [2024]

    Available: https://arxiv.org/abs/2402.12289

    [Online]. Available: https://arxiv.org/abs/2402.12289

    arXiv

Pith tools

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