Pith. sign in

REVIEW 4 major objections 63 references

MobQA: A Benchmark Dataset for Semantic Understanding of Human Mobility Data through Question Answering

T0 review · 4 major / 0 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read MobQA, a 5,800-pair benchmark of GPS-based question answering, shows large language models can retrieve mobility facts but struggle to interpret what the movement means.

desk verdict MobQA abstract looks sensible, but the full text is an unrelated remote-sensing paper, so the submission is incoherent and unverdictable. read the letter →

arxiv 2508.11163 v1 pith:C7E6MFI2 submitted 2025-08-15 cs.CL

Hikaru Asano , Hiroki Ouchi , Akira Kasuga , Ryo Yonetani This is my paper
classification cs.CL
keywords LLMevaluationbenchmarkdatasethumanmobilityGPStrajectoriesquestionansweringsemanticreasoningtrajectorylengthsemantics
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 introduces MobQA, a benchmark of 5,800 question-answer pairs built from human GPS trajectories spanning daily to weekly movement. Its purpose is to measure whether large language models can do more than predict movement — whether they can understand the meaning and reasons behind mobility patterns. The authors claim that current major LLMs handle factual retrieval well but are significantly limited on multiple-choice semantic reasoning and free-form explanation questions, and that longer trajectories make the weakness worse. If the benchmark is valid, it gives the field a reusable instrument and a concrete snapshot of where LLM mobility understanding stands. The full text supplied with this submission is an unrelated remote-sensing paper, so the construction and evaluation controls behind these numbers cannot be checked here.

What carries the argument

MobQA itself — a dataset of 5,800 question-answer pairs over daily-to-weekly GPS trajectories, organized into factual retrieval, multiple-choice reasoning, and free-form explanation items that jointly require spatial, temporal, and semantic reasoning. The benchmark is the instrument that carries the claim: it separates retrieval from inference and interpretation so that LLM performance differences across question types read as differences in semantic understanding.

What would settle it

Take a random subset of the 5,800 pairs, have independent annotators re-create gold answers and score the same model outputs by human judgment; if human scores disagree with the paper's reported ordering across question types, or if inter-annotator agreement on the gold answers is low, the central claim collapses.

Watch

Extended reading notes

Core claim

The paper's central claim is that semantic understanding of mobility data can be decomposed into three layers and tested by question answering: factual retrieval (what happened), multiple-choice reasoning (why it happened), and free-form explanation (what the pattern means). On MobQA, major LLMs show strong performance on the first layer but significant limitations on the second and third, and trajectory length substantially affects effectiveness. The authors present this as evidence that state-of-the-art LLMs have not yet achieved robust semantic mobility understanding.

Load-bearing premise

The central finding stands only if the 5,800 question-answer pairs have correct, unbiased gold answers scored by a reliable, non-contaminated protocol — and the supplied full text is a different paper, so none of these controls can be verified from this submission.

Editorial extensions

If this is right

  • LLMs already handle factual extraction from mobility traces well, so future work should shift toward reasoning and interpretation rather than basic retrieval.
  • Semantic reasoning and free-form explanation are the current bottlenecks in LLM mobility understanding.
  • Trajectory length matters: longer trajectories degrade model effectiveness, so compressing or hierarchically summarizing long GPS traces could improve performance.
  • MobQA provides a reusable evaluation suite for any future model that claims semantic understanding of human mobility data.

Reading between the lines

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

  • If the reported failure is driven by trajectory length, then retrieval-augmented or hierarchical summarization of GPS traces might improve LLM reasoning — a testable extension the paper itself does not run.
  • Human performance on the same 5,800 pairs would calibrate what 'semantic understanding' should mean for mobility and whether the gap the paper reports is a model limitation or a task difficulty artifact.
  • The multiple-choice and free-form layers of MobQA could double as a general probe of spatial-temporal reasoning in LLMs, not just mobility-specific understanding.
  • Because the explanation questions are free-form, the choice of automatic scoring metric likely changes the leaderboard; a validated human-eval subset would be needed before using MobQA as a competitive ranking instrument.
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 / 0 minor

Summary. The submission, as reviewed, consists of an abstract announcing MobQA, a benchmark of 5,800 question-answer pairs over human GPS trajectories, with three question types (factual retrieval, multiple-choice reasoning, free-form explanation) and an evaluation of major LLMs reporting strong factual retrieval but weak reasoning/explanation and a trajectory-length effect. The supplied full text, however, is an entirely different paper: "VFM-Guided Semi-Supervised Detection Transformer under Source-Free Constraints for Remote Sensing Object Detection" (VG-DETR, arXiv:2508.11167v2). The full text contains no mention of MobQA, human mobility, GPS trajectories, question-answer construction, LLM evaluation, or free-form answer scoring. Consequently, every empirical claim in the abstract is unsupported by the reviewed manuscript.

Significance. A well-constructed mobility QA benchmark would be a useful contribution, and the abstract gestures at a concrete resource (a GitHub URL, 5,800 pairs, three question types). However, significance can only be assessed if the benchmark construction and evaluation are described and checkable. The current submission provides no such description, no protocol, no annotation-quality evidence, and no experimental details. The full text is a remote sensing object detection paper, so the benchmark and its findings cannot be evaluated. The paper's potential value is real but entirely unverified in the submitted material.

major comments (4)
  1. [Abstract vs. Full Text] The abstract describes MobQA, LLM evaluation, and trajectory-length effects, but the full text is VG-DETR, a source-free remote sensing object detection paper. The running header identifies arXiv:2508.11167v2, and the body discusses pseudo-labels, DETR, and domain adaptation. There is no overlap of topic. The central claim of the abstract is therefore unsupported by the reviewed text.
  2. [No benchmark construction details] The abstract states that MobQA comprises 5,800 high-quality question-answer pairs, but the manuscript gives no information about how trajectories were collected, how questions were generated, how gold answers were verified, whether expert review or inter-annotator agreement was used, or what quality controls produced the asserted 'high-quality' label. Without this, the benchmark's trustworthiness is unassessable.
  3. [No evaluation protocol] The abstract claims strong factual retrieval but weak reasoning and explanation performance across 'major LLMs.' The full text provides no model list, prompt templates, answer-scoring method (especially for free-form explanations), variance or error bars, or human agreement data. It is therefore impossible to verify the direction or magnitude of the claimed effects, or whether an LLM-based judge might be measuring self-agreement rather than semantic understanding.
  4. [No contamination or representativeness controls] The claim that MobQA measures general LLM semantic understanding of mobility data presupposes that the test items are not memorized from pretraining and that trajectories and questions are representative. The manuscript provides no discussion of contamination checks, no diversity analysis of trajectories, and no evidence that the benchmark supports the broad conclusions stated in the abstract.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation is exhibitable from the MobQA abstract; the supplied full text is an unrelated paper, making the benchmark claims unverifiable but not circular.

full rationale

The claimed contribution of the MobQA abstract is an externally constructed benchmark (5,800 human-authored question-answer pairs over GPS trajectories) plus an evaluation of existing LLMs against that benchmark. There is no derivation chain in which an output is defined in terms of an input, a fitted parameter is renamed as a prediction, or a self-citation is load-bearing. The abstract does not specify how free-form explanations were scored, so one cannot quote any statement that they were graded by the same LLMs under evaluation; flagging that as circular would be speculation. The full text supplied with the submission is a different paper (VG-DETR, arXiv:2508.11167v2) with no mention of MobQA, trajectories, or QA construction. That mismatch prevents checking the benchmark's quality controls, gold-answer consistency, contamination safeguards, or scoring protocol, but missing methodology is a completeness/integrity concern, not a circularity reduction. Under the hard rule that circularity may only be claimed with a quoted reduction or fitted-parameter-as-prediction, no circular step can be identified. Therefore the circularity score is 0.

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

The abstract introduces no fitted parameters and no new physical or theoretical entities. The load-bearing assumptions are all about the benchmark's validity as a measurement instrument: that the questions measure semantic understanding, that the gold answers are correct, and that the free-form answers are scored without circularity. None are verifiable from the abstract, and the submitted full text is unrelated to the abstract, so the ledger cannot be completed.

assumptions (3)
  • domain assumption QA accuracy on MobQA is a valid operationalization of semantic understanding of mobility data
    The abstract equates benchmark question answering with semantic interpretation of trajectories, a construct assumption that underpins the entire evaluation.
  • domain assumption The 5,800 question-answer pairs have correct gold answers and were quality-controlled
    The evaluation results are only meaningful if the gold answers are correct and consistent; the abstract reports no inter-annotator agreement or review procedure.
  • domain assumption Free-form explanation answers are scored reliably and fairly
    Free-form scoring is the hardest part of the benchmark; if it uses the same LLMs under evaluation or an unvalidated metric, the explanation results would be confounded.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MobQA: A Benchmark Dataset for Semantic Understanding of Human Mobility Data through Question Answering." pith.science (2026). https://pith.science/paper/C7E6MFI2

@misc{pith2026250811163,
  author       = {Pith},
  title        = {Pith review of: MobQA: A Benchmark Dataset for Semantic Understanding of Human Mobility Data through Question Answering},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/C7E6MFI2}},
  note         = {Machine review of arXiv:2508.11163}
}
read the original abstract

This paper presents MobQA, a benchmark dataset designed to evaluate the semantic understanding capabilities of large language models (LLMs) for human mobility data through natural language question answering. While existing models excel at predicting human movement patterns, it remains unobvious how much they can interpret the underlying reasons or semantic meaning of those patterns. MobQA provides a comprehensive evaluation framework for LLMs to answer questions about diverse human GPS trajectories spanning daily to weekly granularities. It comprises 5,800 high-quality question-answer pairs across three complementary question types: factual retrieval (precise data extraction), multiple-choice reasoning (semantic inference), and free-form explanation (interpretive description), which all require spatial, temporal, and semantic reasoning. Our evaluation of major LLMs reveals strong performance on factual retrieval but significant limitations in semantic reasoning and explanation question answering, with trajectory length substantially impacting model effectiveness. These findings demonstrate the achievements and limitations of state-of-the-art LLMs for semantic mobility understanding.\footnote{MobQA dataset is available at https://github.com/CyberAgentAILab/mobqa.}

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

63 extracted references · 54 canonical work pages

  1. [1]

    Instance Feature Visualization: We evaluate the sim- ilarity between the reference regions (marked with yellow stars) and all other spatial locations in the feature maps extracted by DINOv2, as shown in Fig. 4. It is evident that TABLE VIII DETECTION PERFORMANCE WITH DIFFERENT VFM S ACROSS MULTIPLE SCENARIOS . VFM xView→DOTA SRSD→DIOR HRRSD→SSDD 1% 5% 10%...

    work page

  2. [3]

    Zhu et al

    proposed the first UDA approach built on the one-stage detector RetinaNet [31] and introduced an adaptive feature alignment strategy to address the challenge of accurately aligning sparse foreground objects. Zhu et al. [32] incorporated a mean-teacher framework into remote sensing adaptation, using dual detection heads to suppress biased information and r...

    work page

  3. [4]

    Zhang et al

    were the first to propose a source-free object detector and to develop a self-entropy based approach for setting an appropriate confidence threshold. Zhang et al. [5] constructed multiple prototypes for instance features within a Faster R- CNN detector and leveraged the consistency between teacher- and student-prototype distributions to correct the genera...

    work page

  4. [5]

    The Effect of Different Vision Foundation Models on Feature Extraction: We further evaluate the performance of our method with different VFMs across three cross-domain scenarios, as reported in Table VIII. DINOv3 [69] outperforms DINOv2 in optical imagery, owing to its ability to capture finer-grained feature representations that are particularly ben- efi...

    work page

  5. [8]

    VFM-Guided Semi-Supervised Detection Transformer under Source-Free Constraints for Remote Sensing Object Detection

    and feature alignment techniques [1], [9]–[11] are generally infeasible. Instead, self-training paradigms based on the mean teacher framework [12] have become the dominant approach. These paradigms consist of teacher and student models with identical structures, where the teacher model generates pseudo- labels to guide the student model in supervised trai...

    work page Pith review arXiv 2025

  6. [12]

    Unbiased teacher for semi-supervised object detection,

    Y .-C. Liu, C.-M. Ma, Z. He, C.-W. Kuo, K. Chen, P. Zhang, B. Wu, Z. Kira, and P. Vajda, “Unbiased teacher for semi-supervised object detection,” in Proceedings of the International Conference on Learning Representations, 2021

    work page 2021

  7. [13]

    Ema: A process model of appraisal dynamics,

    S. C. Marsella and J. Gratch, “Ema: A process model of appraisal dynamics,” Cognitive Systems Research , vol. 10, no. 1, p. 70–90, Mar 2009

    work page 2009

  8. [14]

    Periodically exchange teacher-student for source-free object detection,

    Q. Liu, L. Lin, Z. Shen, and Z. Yang, “Periodically exchange teacher-student for source-free object detection,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023, pp. 6414–6424

    work page 2023

Show all 63 references
  1. [15]

    Dynamic retraining-updating mean teacher for source-free object detection,

    T. L. B. Khanh, H.-H. Nguyen, L. H. Pham, D. N.-N. Tran, and J. W. Jeon, “Dynamic retraining-updating mean teacher for source-free object detection,” in Proceedings of the European Conference on Computer Vision. Springer, 2024, pp. 328–344

    2024

  2. [16]

    Dinov2: Learning robust visual features without supervision,

    M. Oquab, T. Darcet, T. Moutakanni, H. V o, M. Szafraniec, V . Khalidov, P. Fernandez, D. Haziza, F. Massa, A. El-Nouby et al. , “Dinov2: Learning robust visual features without supervision,” arXiv preprint arXiv:2304.07193, 2023

    2023 arXiv

  3. [17]

    Segment anything,

    A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y . Loet al., “Segment anything,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 4015–4026

    2023

  4. [18]

    Grounding dino: Marrying dino with grounded pre- training for open-set object detection,

    S. Liu, Z. Zeng, T. Ren, F. Li, H. Zhang, J. Yang, Q. Jiang, C. Li, J. Yang, H. Su et al. , “Grounding dino: Marrying dino with grounded pre- training for open-set object detection,” in Proceedings of the European Conference on Computer Vision . Springer, 2024, pp. 38–55

    2024

  5. [19]

    Remoteclip: A vision language foundation model for remote sensing,

    F. Liu, D. Chen, Z. Guan, X. Zhou, J. Zhu, Q. Ye, L. Fu, and J. Zhou, “Remoteclip: A vision language foundation model for remote sensing,” IEEE Transactions on Geoscience and Remote Sensing , vol. 62, pp. 1– 16, 2024

    2024

  6. [20]

    Some methods for classification and analysis of multi- variate observations,

    J. MacQueen, “Some methods for classification and analysis of multi- variate observations,” in Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability, Volume 1: Statistics , vol. 5. University of California press, 1967, pp. 281–298

    1967

  7. [21]

    Sinkhorn distances: Lightspeed computation of optimal transport,

    M. Cuturi, “Sinkhorn distances: Lightspeed computation of optimal transport,” Advances in Neural Information Processing Systems , vol. 26, 2013

    2013

  8. [22]

    xview: Objects in context in overhead imagery,

    D. Lam, R. Kuzma, K. McGee, S. Dooley, M. Laielli, M. Klaric, Y . Bulatov, and B. McCord, “xview: Objects in context in overhead imagery,” arXiv preprint arXiv:1802.07856 , 2018

    2018 arXiv

  9. [23]

    Dota: A large-scale dataset for object detection in aerial images,

    G.-S. Xia, X. Bai, J. Ding, Z. Zhu, S. Belongie, J. Luo, M. Datcu, M. Pelillo, and L. Zhang, “Dota: A large-scale dataset for object detection in aerial images,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2018, pp. 3974–3983

    2018

  10. [24]

    Object detection in optical remote sensing images: A survey and a new benchmark,

    K. Li, G. Wan, G. Cheng, L. Meng, and J. Han, “Object detection in optical remote sensing images: A survey and a new benchmark,” ISPRS Journal of Photogrammetry and Remote Sensing , p. 296–307, Jan 2020

    2020

  11. [25]

    Hierarchical and robust convolutional neural network for very high-resolution remote sensing object detection,

    Y . Zhang, Y . Yuan, Y . Feng, and X. Lu, “Hierarchical and robust convolutional neural network for very high-resolution remote sensing object detection,” IEEE Transactions on Geoscience and Remote Sens- ing, vol. 57, no. 8, pp. 5535–5548, 2019

    2019

  12. [26]

    Ship detection in sar images based on an improved faster r-cnn,

    J. Li, C. Qu, and J. Shao, “Ship detection in sar images based on an improved faster r-cnn,” in Proceedings of the 2017 SAR in Big Data Era: Models, Methods and Applications . IEEE, 2017, pp. 1–6

    2017

  13. [27]

    Faster r-cnn: Towards real-time object detection with region proposal networks,

    S. Ren, K. He, R. Girshick, and J. Sun, “Faster r-cnn: Towards real-time object detection with region proposal networks,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 39, no. 6, pp. 1137– 1149, 2017

    2017

  14. [28]

    End-to-end object detection with transformers,

    N. Carion, F. Massa, G. Synnaeve, N. Usunier, A. Kirillov, and S. Zagoruyko, “End-to-end object detection with transformers,” in Proceedings of the European Conference on Computer Vision. Springer, 2020, pp. 213–229

    2020

  15. [29]

    Exploring sequence feature alignment for domain adaptive detection transformers,

    W. Wang, Y . Cao, J. Zhang, F. He, Z.-J. Zha, Y . Wen, and D. Tao, “Exploring sequence feature alignment for domain adaptive detection transformers,” in Proceedings of the ACM International Conference on Multimedia, 2021, pp. 1730–1738

    2021

  16. [30]

    Da- detr: Domain adaptive detection transformer with information fusion,

    J. Zhang, J. Huang, Z. Luo, G. Zhang, X. Zhang, and S. Lu, “Da- detr: Domain adaptive detection transformer with information fusion,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 23 787–23 798

    2023

  17. [31]

    Focal loss for dense object detection,

    T.-Y . Lin, P. Goyal, R. Girshick, K. He, and P. Doll ´ar, “Focal loss for dense object detection,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2017, pp. 2980–2988

    2017

  18. [32]

    Dualda-net: Dual-head rectification for cross-domain object detection of remote sensing,

    Y . Zhu, X. Sun, W. Diao, H. Wei, and K. Fu, “Dualda-net: Dual-head rectification for cross-domain object detection of remote sensing,” IEEE Transactions on Geoscience and Remote Sensing , vol. 61, pp. 1–16, 2023

    2023

  19. [33]

    Remote sensing teacher: Cross-domain detection transformer with learnable frequency- enhanced feature alignment in remote sensing imagery,

    J. Han, W. Yang, Y . Wang, L. Chen, and Z. Luo, “Remote sensing teacher: Cross-domain detection transformer with learnable frequency- enhanced feature alignment in remote sensing imagery,” IEEE Trans- actions on Geoscience and Remote Sensing , vol. 62, no. 5619814, pp. 1–14, 2024

    2024

  20. [34]

    Balanced teacher for source-free ob- ject detection,

    J. Deng, W. Li, and L. Duan, “Balanced teacher for source-free ob- ject detection,” IEEE Transactions on Circuits and Systems for Video Technology, 2024

    2024

  21. [35]

    Dense learning based semi-supervised object detection,

    B. Chen, P. Li, X. Chen, B. Wang, L. Zhang, and X.-S. Hua, “Dense learning based semi-supervised object detection,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2022, pp. 4815–4824

    2022

  22. [36]

    Dense teacher: Dense pseudo-labels for semi-supervised object detection,

    H. Zhou, Z. Ge, S. Liu, W. Mao, Z. Li, H. Yu, and J. Sun, “Dense teacher: Dense pseudo-labels for semi-supervised object detection,” in Proceedings of the European Conference on Computer Vision. Springer, 2022, pp. 35–50

    2022

  23. [37]

    Efficient non-maximum suppression,

    A. Neubeck and L. Van Gool, “Efficient non-maximum suppression,” in Proceedings of the International Conference on Pattern Recognition , 2006, pp. 850–855

    2006

  24. [38]

    End-to-end semi-supervised object detection with soft teacher,

    M. Xu, Z. Zhang, H. Hu, J. Wang, L. Wang, F. Wei, X. Bai, and Z. Liu, “End-to-end semi-supervised object detection with soft teacher,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 3060–3069

    2021

  25. [39]

    Dual teacher: Improv- ing the reliability of pseudo labels for semi-supervised oriented object detection,

    Z. Fang, J. Ren, J. Zheng, R. Chen, and H. Zhao, “Dual teacher: Improv- ing the reliability of pseudo labels for semi-supervised oriented object detection,” IEEE Transactions on Geoscience and Remote Sensing, 2024

    2024

  26. [40]

    Minimizing sample redundancy for label-efficient object detection in aerial images,

    R. Zhang, C. Xu, H. Zhu, F. Xu, W. Yang, H. Zhang, and G.-S. Xia, “Minimizing sample redundancy for label-efficient object detection in aerial images,” IEEE Transactions on Geoscience and Remote Sensing , vol. 63, pp. 1–14, 2025

    2025

  27. [41]

    Omni-detr: Omni-supervised object detection with transformers,

    P. Wang, Z. Cai, H. Yang, G. Swaminathan, N. Vasconcelos, B. Schiele, and S. Soatto, “Omni-detr: Omni-supervised object detection with transformers,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2022, pp. 9367–9376

    2022

  28. [42]

    Semi-detr: Semi-supervised object detection with detection transformers,

    J. Zhang, X. Lin, W. Zhang, K. Wang, X. Tan, J. Han, E. Ding, J. Wang, and G. Li, “Semi-detr: Semi-supervised object detection with detection transformers,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 23 809–23 818

    2023

  29. [43]

    Sparse semi-detr: sparse learnable queries for semi-supervised object detection,

    T. Shehzadi, K. A. Hashmi, D. Stricker, and M. Z. Afzal, “Sparse semi-detr: sparse learnable queries for semi-supervised object detection,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 5840–5850

    2024

  30. [44]

    Detect everything with few examples,

    X. Zhang, Y . Liu, Y . Wang, and A. Boularias, “Detect everything with few examples,” arXiv preprint arXiv:2309.12969 , 2023

    2023 arXiv

  31. [45]

    Cross-domain few-shot object detection via enhanced open-set object detector,

    Y . Fu, Y . Wang, Y . Pan, L. Huai, X. Qiu, Z. Shangguan, T. Liu, Y . Fu, L. Van Gool, and X. Jiang, “Cross-domain few-shot object detection via enhanced open-set object detector,” in Proceedings of the European Conference on Computer Vision . Springer, 2024, pp. 247–264

    2024

  32. [46]

    Large self-supervised models bridge the gap in domain adaptive object detection,

    M.-A. Lavoie, A. Mahmoud, and S. L. Waslander, “Large self-supervised models bridge the gap in domain adaptive object detection,” in Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2025, pp. 4692–4702

    2025

  33. [47]

    Frozen- detr: Enhancing detr with image understanding from frozen foundation models,

    S. Fu, J. Yan, Q. Yang, X. Wei, X. Xie, and W.-S. Zheng, “Frozen- detr: Enhancing detr with image understanding from frozen foundation models,” arXiv preprint arXiv:2410.19635 , 2024

    2024 arXiv

  34. [48]

    Good: Towards domain generalized oriented object detection,

    Q. Bi, B. Zhou, J. Yi, W. Ji, H. Zhan, and G.-S. Xia, “Good: Towards domain generalized oriented object detection,” ISPRS Journal of Pho- togrammetry and Remote Sensing , vol. 223, pp. 207–220, 2025

    2025

  35. [49]

    Learning transferable visual models from natural language supervision,

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark et al., “Learning transferable visual models from natural language supervision,” in Proceedings of the International Conference on Machine Learning, pages=8748–8763, ye...

    2021

  36. [50]

    Dino: Detr with improved denoising anchor boxes for end-to- end object detection,

    H. Zhang, F. Li, S. Liu, L. Zhang, H. Su, J. Zhu, L. M. Ni, and H.-Y . Shum, “Dino: Detr with improved denoising anchor boxes for end-to- end object detection,” in Proceedings of the International Conference on Learning Representations , 2023

    2023

  37. [51]

    An image is worth 16x16 words: Transformers for image recognition at scale,

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby, “An image is worth 16x16 words: Transformers for image recognition at scale,” in Proceedings of the International Con...

    2021

  38. [52]

    Exploring robust features for few-shot object detection in satellite 15 imagery,

    X. Bou, G. Facciolo, R. G. V on Gioi, J.-M. Morel, and T. Ehret, “Exploring robust features for few-shot object detection in satellite 15 imagery,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 430–439

    2024

  39. [53]

    Segment any change,

    Z. Zheng, Y . Zhong, L. Zhang, and S. Ermon, “Segment any change,” Advances in Neural Information Processing Systems , vol. 37, pp. 81 204–81 224, 2024

    2024

  40. [54]

    Evaluating general purpose vision foundation models for medical image analysis: An experimental study of dinov2 on radiology benchmarks,

    M. Baharoon, W. Qureshi, J. Ouyang, Y . Xu, A. Aljouie, and W. Peng, “Evaluating general purpose vision foundation models for medical image analysis: An experimental study of dinov2 on radiology benchmarks,” arXiv preprint arXiv:2312.02366 , 2023

    2023 arXiv

  41. [55]

    Dino-reg: General purpose image encoder for training-free multi-modal deformable medical image registration,

    X. Song, X. Xu, and P. Yan, “Dino-reg: General purpose image encoder for training-free multi-modal deformable medical image registration,” in International Conference on Medical Image Computing and Computer- Assisted Intervention. Springer, 2024, pp. 608–617

    2024

  42. [56]

    Mask r-cnn,

    K. He, G. Gkioxari, P. Doll ´ar, and R. Girshick, “Mask r-cnn,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2017, pp. 2961–2969

    2017

  43. [57]

    Datr: Unsupervised domain adaptive detection transformer with dataset-level adaptation and prototypical alignment,

    L. Chen, J. Han, and Y . Wang, “Datr: Unsupervised domain adaptive detection transformer with dataset-level adaptation and prototypical alignment,” IEEE Transactions on Image Processing , vol. 34, pp. 982– 994, 2025

    2025

  44. [58]

    Every pixel matters: Center-aware feature alignment for domain adaptive object detector,

    C. Hsu, Y .-H. Tsai, Y .-Y . Lin, and M.-H. Yang, “Every pixel matters: Center-aware feature alignment for domain adaptive object detector,” in Proceedings of the European Conference on Computer Vision, 2020, pp. 733–748

    2020

  45. [59]

    Instance relation graph guided source- free domain adaptive object detection,

    V . VS, P. Oza, and V . M. Patel, “Instance relation graph guided source- free domain adaptive object detection,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 3520–3530

    2023

  46. [60]

    Enhanc- ing source-free domain adaptive object detection with low-confidence pseudo label distillation,

    I. Yoon, H. Kwon, J. Kim, J. Park, H. Jang, and K. Sohn, “Enhanc- ing source-free domain adaptive object detection with low-confidence pseudo label distillation,” in Proceedings of the European Conference on Computer Vision . Springer, 2024, pp. 337–353

    2024

  47. [61]

    Unbiased teacher v2: Semi-supervised object detection for anchor-free and anchor-based detectors,

    Y .-C. Liu, C.-Y . Ma, and Z. Kira, “Unbiased teacher v2: Semi-supervised object detection for anchor-free and anchor-based detectors,” in Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 9819–9828

    2022

  48. [62]

    Multi-clue consistency learning to bridge gaps between general and oriented object in semi-supervised detection,

    C. Wang, C. Xu, X. Li, Y . Li, X. Guo, Z. Gu, and Z. Cui, “Multi-clue consistency learning to bridge gaps between general and oriented object in semi-supervised detection,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 39, no. 7, 2025, pp. 7582–7590

    2025

  49. [63]

    Rareplanes: Synthetic data takes flight,

    J. Shermeyer, T. Hossler, A. V . Etten, D. Hogan, R. Lewis, and D. Kim, “Rareplanes: Synthetic data takes flight,” in Proceedings of the IEEE Winter Conference on Applications of Computer Vision , Jan 2021. [Online]. Available: http://dx.doi.org/10.1109/wacv48630.2021.00025

    2021

  50. [64]

    Unsupervised domain adaptation for remote- sensing vehicle detection using domain-specific channel recalibration,

    W. Liu, J. Liu, and B. Luo, “Unsupervised domain adaptation for remote- sensing vehicle detection using domain-specific channel recalibration,” IEEE Geoscience and Remote Sensing Letters , vol. 20, pp. 1–5, 2023

    2023

  51. [65]

    Hierarchical similarity alignment for domain adaptive ship detection in sar images,

    J. Zhang, S. Li, Y . Dong, B. Pan, and Z. Shi, “Hierarchical similarity alignment for domain adaptive ship detection in sar images,” IEEE Transactions on Geoscience and Remote Sensing , vol. 60, pp. 1–11, 2022

    2022

  52. [66]

    Fsda-detr: Few- shot domain-adaptive object detection transformer in remote sensing im- agery,

    B. Yang, J. Han, X. Hou, D. Zhou, W. Liu, and F. Bi, “Fsda-detr: Few- shot domain-adaptive object detection transformer in remote sensing im- agery,” IEEE Transactions on Geoscience and Remote Sensing , vol. 63, pp. 1–16, 2025

    2025

  53. [67]

    Ima- genet: A large-scale hierarchical image database,

    J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei, “Ima- genet: A large-scale hierarchical image database,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2009, pp. 248–255

    2009

  54. [68]

    Adam: A method for stochastic optimization,

    D. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980 , 2014

    2014 arXiv

  55. [69]

    Sim ´eoni, H

    O. Sim ´eoni, H. V . V o, M. Seitzer, F. Baldassarre, M. Oquab, C. Jose, V . Khalidov, M. Szafraniec, S. Yi, M. Ramamonjisoa et al., “Dinov3,” arXiv preprint arXiv:2508.10104 , 2025

    2025 arXiv

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.