Pith. sign in

REVIEW 3 major objections 4 minor 97 references

Filling Before Advancing: Capability-Gap-Driven Post-Training for Scenario-Specialized Remote Sensing MLLMs

T0 review · 3 major / 4 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read Staged post-training that fills prerequisite capability gaps before scenario-specific tuning improves harbor understanding in remote-sensing MLLMs.

desk verdict A carefully ablated staged-fine-tuning recipe with genuinely new artifacts, but the headline ordering claim rests on a private, author-built benchmark and an unmatched data budget—send to review and require independent validation. read the letter →

arxiv 2607.22205 v2 pith:SN2RFZCK submitted 2026-07-24 cs.CV cs.AI

Yuheng Zong , Minghua Wang , Xin Zhao , Zhi-Hui Zhan , Antonio Plaza , Jon Atli Benediktsson This is my paper · ORCID
classification cs.CVcs.AI
keywords remotesensingmultimodallargelanguagemodelsscenariospecializationpost-traininginstructiontuningcapabilitygapcoastalharborunderstandingsupervisedfine-tuningmulti-source
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 argues that specializing a remote-sensing multimodal large language model to a fine-grained scenario such as coastal harbor understanding should not be done by single-stage supervised fine-tuning on target-domain samples. It proposes a three-stage route—'filling before advancing'—that first anchors overhead visual-language alignment, then converges shared multi-sensor priors across target and bridging scenes, and only then tunes for evidence-grounded scenario behavior. The authors build CPRS, a three-layer supervision dataset that assigns scarce expert supervision to those separable roles, and HarborEval, an eight-track diagnostic benchmark spanning RGB, SAR, PAN, and NIR. Under matched training budgets, the ordered route raises HarborEval from 57.95 to 70.29 on LLaVA-v1.5 and from 81.09 to 83.37 on Qwen3-VL, and also beats collapsed single-stage alternatives and existing RS-MLLMs. The point is that ordering and role assignment of scarce supervision, not just data volume, is what drives scenario specialization.

What carries the argument

The load-bearing object is the ordered three-stage route coupled to CPRS's three supervision layers: RS-Anchor (D1), Bridge-Conv (D2), and Scenario-EG (D3). Each stage initializes from the previous model and consumes one supervision layer, so the training trajectory is a progressive narrowing from broad RGB RS semantics, through multi-source bridge-domain convergence (with staged multi-teacher distillation producing verified SFT records), to final evidence-grounded scenario tuning. HarborEval, an eight-track diagnostic benchmark, is the measurement instrument that converts the capability claims into a single macro-averaged score; its tracks are tied to capability roles (object/zone recogniti

What would settle it

Re-implement FBA, Direct-SFT, and Collapsed-SFT from the released CPRS pools and re-score all models on HarborEval with independently annotated answer keys and a different judge; if Direct-SFT matches or beats FBA under independent scoring, or if the ordering advantage disappears when D3 alone is given the same total compute budget as the full route, the central claim is falsified.

Watch

Extended reading notes

Core claim

FBA's central claim is that the capability gaps between a general MLLM and a scenario-specialized RS-MLLM—overhead-view semantics, sensor-aware observability, target-bridging context, evidence grounding, and calibrated rejection—can be closed progressively by an ordered route in which each stage has one separable role. Concretely: S1 adapts the model on 569,853 RGB image-caption pairs for RS semantic anchoring; S2 trains on 187,296 multi-source SFT samples mixing target harbor scenes with bridging coastal-port scenes across RGB, SAR, PAN, and NIR to consolidate shared RS priors; S3 tunes on 53,000 evidence-grounded harbor instruction samples with negatives and uncertainty labels. The paper r

Load-bearing premise

The central measurement assumption is that HarborEval—with its private answer keys, accepted-answer sets, and judge rubric—is a valid and unbiased measure of harbor-scenario specialization; if those private answers or rubrics encode exactly the behaviors FBA was trained to produce, the reported ordering gains could be an artifact of evaluation design.

Editorial extensions

If this is right

  • If FBA is right, scarcity of high-quality scenario data is not itself the binding constraint; the ordering and functional role of available supervision is what determines specialization quality.
  • The route should transfer to other remote-sensing vertical scenarios wherever one can identify bridging scenes and evidence-grounding needs.
  • Replacing any stage with generic supervision (generic image-text at S1, non-bridging data at S2, or non-evidence-grounded data at S3) degrades downstream HarborEval, so each layer's role is load-bearing.
  • A smaller, well-ordered 810K-sample route can beat larger single-stage instruction-tuned models on scenario-specific diagnostics.
  • Because the route is budget-matched, the reported gains are attributable to ordering and composition, not to additional compute or data.

Reading between the lines

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

  • Editorial inference: the paradigm predicts that the same three-role ordering will show smaller gains for scenarios where the base model already has strong priors, and larger gains for scenarios far from natural-image pretraining; this is testable by applying FBA to scenarios with known pretraining distance.
  • Editorial inference: because the headline comparisons rely on a self-constructed benchmark with private answer keys, independent re-scoring with released data and an external judge is needed before the ordering claim is treated as settled.
  • Editorial inference: a natural extension is to make stage boundaries adaptive—deciding when anchoring is sufficient by monitoring intermediate diagnostics rather than fixing one epoch per stage—which the current frozen-manifest design deliberately avoids.
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

3 major / 4 minor

Summary. The paper casts scenario specialization of remote-sensing MLLMs as a capability-gap-driven post-training problem and proposes FBA (filling before advancing), an ordered three-stage route: S1 RS semantic anchoring on large-scale RGB image-caption data, S2 domain-bridge convergence on multi-source target/bridging SFT data, and S3 evidence-grounded scenario tuning on harbor-specific instruction data. The authors construct CPRS, a three-layer dataset, and HarborEval, an eight-track diagnostic benchmark, and report that FBA outperforms Direct-SFT and Collapsed-SFT on LLaVA-v1.5 and Qwen3-VL, while also leading on harbor-related VRSBench/RSVQA subsets and an expert-scored OpenEval. The paper includes extensive leakage audits, role-replacement ablations, and stage-wise analyses intended to support the claim that the order of capability filling matters.

Significance. If the ordering claim is accepted, the paper provides a useful framework for scenario-specialized RS-MLLMs under scarce high-quality supervision. The internal controls are a clear strength: the role-replacement ablations (Table 4), the S1+Collapsed comparison (Table 1), the source-record holdout, and the field-level public/private separation (S3.2–S3.5) are unusually careful. The work also ships public examples and code/weight release plans. However, the central evidence rests on a self-constructed benchmark whose task formats, answer vocabulary, and scoring rubrics closely mirror the final training layer, and all headline comparisons are single-run point estimates. The external validation is limited to author-selected public-data subsets and an author-designed expert protocol, so the construct validity of HarborEval is the decisive risk.

major comments (3)
  1. [Abstract / Table 1 / S4 (training budgets)] The claim of improvement 'under comparable training budgets' is not supported by the setup. Direct-SFT uses only D3 (53,000 samples, ~3.3K optimizer steps at batch 16), while FBA uses D1 (569,853 samples, ~35.6K steps), D2 (187,296 samples, ~11.7K steps), and D3 (~3.3K steps), for roughly 50K total steps. Even Collapsed-SFT uses only D2∪D3 (~15K steps). The headline gains (57.95→70.29 on LLaVA; 81.09→83.37 on Qwen3-VL) therefore conflate the ordered-route hypothesis with extra data and compute. Please rerun Direct-SFT with matched data/compute budgets (e.g., adding an equivalent volume of generic/RS data to D3), or explicitly report FLOPs/steps and rephrase the claim.
  2. [S3.2–S3.5, Table S4, Tables 1 and 2] HarborEval's construct validity is the main load-bearing concern. D3 and HarborEval share the same task taxonomy (presence validation, relation reasoning, grid localization, functional zones, open rejection), the same Yes/No/Cannot-determine vocabulary, the same 3×3 grid conventions, and the same evidence-grounding/forbidden-claim rules. The source-record holdout prevents image-level overlap but does not prevent format/rubric conformance: a model trained on D3 can raise HarborEval by internalizing exactly the behaviors HarborEval rewards. The public VRSBench/RSVQA subsets are author-selected 'harbor-related' slices scored with author-chosen protocols, and Table 2 compares against published checkpoints rather than rerun Direct/Collapsed baselines on those external sets. Please provide an independent evaluation (full public benchmark or third-party annotation) or release the private answer
  3. [Tables 1, 3, 4 / S4] All reported numbers appear to be single-run point estimates with no error bars, confidence intervals, or multiple seeds, despite the paper's language of 'consistently outperforms.' Several supporting differences are small (e.g., Qwen3-VL overall 83.37 vs 81.09; Relation 83.15 vs 81.46; Grid 68.61 vs 66.06). In addition, the route uses multiple free hyperparameters (stage-wise learning rates, LoRA ranks, data mixture sizes, track aggregation weights) and no sensitivity analysis is reported. Please report at least 3 seeds with variance, or explicitly justify why single-seed results are sufficient, and provide a sensitivity check on the most influential stage hyperparameters.
minor comments (4)
  1. [Table 3] For Qwen3-VL, RS-VL Val. decreases after S2 and S3 (95.29 → 93.37 → 92.22). The text says intermediate diagnostics 'remain robust as supervision becomes progressively focused,' which is true in magnitude but not monotonic; please clarify whether this decline is expected and how it should be interpreted.
  2. [S3.2] The source-record holdout is described as distinct from field-level controls. It would help to state explicitly whether HarborEval images and CPRS D3 images are drawn from disjoint geographic regions/source families, or only disjoint at the exact-record level, since geographic similarity could still present a contamination channel.
  3. [S5.4 / OpenEval] No inter-annotator agreement or adjudication statistics are reported for the expert OpenEval scoring. Given that OpenEval is one of the few external-looking validations, please report agreement metrics or at least the number of adjudicated cases.
  4. [Table 2] The 'Data Scale' column mixes different training corpora and pretraining bases, and some baselines (GeoChat) inherit large general-purpose samples. Consider adding a column for the base model and noting that the comparison is indicative, not matched.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the route-ordering claim is supported by same-data ablations, role-replacement controls, and public external benchmarks; the author-built HarborEval alignment is a benchmark-validity caveat, not a derivation-level reduction.

full rationale

The paper's central claim is an empirical comparison of post-training routes, not a derivation from an assumed theorem. No equation in the paper defines the predicted HarborEval outcome in terms of its inputs: Eq. (2) merely formalizes sequential adaptation, Eq. (4) macro-averages scores, and the supplementary formulas are scoring metrics. The main FBA-vs-baseline result is produced by controlled experiments in which FBA, Collapsed-SFT, and S1+Collapsed use the same underlying CPRS supervision layers (D1, D2, D3) and differ only in ordering; Table 1 shows FBA outperforms Collapsed-SFT with the same data (70.29 vs 55.74 on LLaVA-v1.5; 83.37 vs 79.36 on Qwen3-VL). Table 4's role-replacement controls additionally show each supervision layer contributes beyond the others, which would be very unlikely if the result were forced by construction. External grounding exists: FBA also leads on harbor-related subsets of the public VRSBench/RSVQA test splits and on expert-scored OpenEval (Table 2), and the evaluation records are held out from training manifests (S3.2-S3.3). The strongest potential concern is that HarborEval was built by the same authors and shares the task taxonomy and evidence conventions used in D3; the paper itself acknowledges the alignment, e.g., S7 states the curation policy is designed to 'align the training language with the rejection behavior evaluated by HarborEval,' and S3.9 lists remaining benchmark boundaries. However, this is a benchmark-validity / train-eval-conformance caveat, not a circular reduction: the HarborEval answers are private and held out, the same scoring protocol is applied identically to all routes, and the route-ordering advantage survives on public benchmark subsets and against external checkpoints. There is no fitted parameter renamed as a prediction, no load-bearing self-citation, and no imported uniqueness theorem. The result may be questioned on construct validity or benchmark independence grounds, but it does not reduce by construction to its inputs.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The paper contributes an empirical training recipe and two new data artifacts; the scientific claim rests on several hand-chosen design choices and self-constructed measurement, so the ledger is dominated by domain assumptions rather than free parameters in the physics sense.

free parameters (4)
  • Stage data mixture sizes (D1, D2, D3) = 569,853 / 187,296 / 53,000 samples; D3 modality mix RGB 70.1%, SAR 23.7%, NIR 2.2%, PAN 4.0%
    Hand-chosen to define the progressive curriculum; the 15%/55%/85% harbor-progression and non-RGB proportions are design choices, not derived from theory.
  • Stage-wise learning rates and LoRA ranks = LLaVA: S1 1e-3 (projector), S2 2e-4/1e-4, S3 1e-4; Qwen3-VL: S1 1e-4, S2 3e-5, S3 3e-5; LoRA rank 64/alpha 128
    Hand-selected hyperparameters; central results could shift with different schedules or adapter settings.
  • Track aggregation weights = HarborEval unweighted macro over 8 tracks; MultiSource Val omega=(0.20,0.25,0.20,0.20,0.10,0.05)
    Headline scores are defined by these weights; changing them can change route rankings.
  • Soft grid similarity weights = same cell 1.0, edge-adjacent 0.5, diagonal 0.25, else 0
    T4 grid-grounding scoring protocol chosen by hand; influences the spatial-understanding track scores.
assumptions (5)
  • ad hoc to paper Capability gaps for scenario specialization are separable and can be filled in a fixed ordered sequence (S1 then S2 then S3).
    Core premise of FBA; no independent evidence that the three stages map onto disjoint capability layers beyond the authors' own diagnostics.
  • domain assumption Bridge-Conv samples share transferable RS priors with harbor scenes that improve downstream specialization.
    Motivated by post-hoc embedding similarity (cosine 0.92 to harbor centroid, Fig. 4); the 'SharedRS' equation (Eq. 3) is conceptual, not operationalized.
  • domain assumption LLM-teacher-generated and LLM-verifier-approved instruction data (SMT) is high-quality, grounded supervision.
    Stage 2's 187,296 samples and parts of Stage 3 are produced or rewritten by Tmeta/Tsyn/Tver; quality rests on teacher/verifier reliability rather than large-scale human annotation.
  • domain assumption HarborEval private answers, accepted-answer sets, and rubrics are correct and free of training contamination.
    All headline route comparisons use this self-constructed benchmark; the leakage audit (S3.2–S3.3) is internal, not independently verifiable.
  • domain assumption The fixed judge (Doubao-Seed-1.8-Vision) and expert scoring faithfully measure grounded reporting.
    T7, MultiSource Val., and OpenEval rely on judge or expert scoring; no inter-rater reliability or judge calibration is reported.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Filling Before Advancing: Capability-Gap-Driven Post-Training for Scenario-Specialized Remote Sensing MLLMs." pith.science (2026). https://pith.science/paper/SN2RFZCK

@misc{pith2026260722205,
  author       = {Pith},
  title        = {Pith review of: Filling Before Advancing: Capability-Gap-Driven Post-Training for Scenario-Specialized Remote Sensing MLLMs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SN2RFZCK}},
  note         = {Machine review of arXiv:2607.22205}
}
read the original abstract

Remote sensing multimodal large language models (RS-MLLMs) have improved general aerial-image understanding. However, Earth observation applications require fine-grained scenario specialization, constrained by scarce high-quality scenario data and incomplete capability coverage. We formulate this adaptation as a capability-gap-driven post-training problem and propose filling before advancing (FBA). Rather than relying on single-stage supervised fine-tuning (SFT) over target-domain samples, FBA first fills prerequisite capability gaps before advancing toward scenario specialization. We instantiate FBA for coastal harbor understanding, a representative multi-source scenario, by constructing CPRS (Coastal-Port Remote Sensing), a three-layer supervision dataset coupled with three ordered stages: (1) RS semantic anchoring for overhead-view visual-language alignment; (2) domain-bridge convergence for shared RS priors across target and bridging scenarios under different modalities; and (3) evidence-grounded scenario tuning for downstream performance. We construct HarborEval, an eight-track diagnostic benchmark covering perception, spatial understanding, robustness, and generation. Under comparable training budgets, HarborEval increases from 57.95 with Direct-SFT to 70.29 with FBA on LLaVA-v1.5, and from 81.09 to 83.37 on Qwen3-VL. FBA also outperforms Collapsed-SFT and leads on harbor-related VRSBench/RSVQA subsets and OpenEval. Stage-wise and role-replacement analyses validate progressive gap filling and stage-specific roles. Public examples and release updates for CPRS, HarborEval, code, and trained weights are available at https://github.com/Z0ngL1ng/filling-before-advancing.

Figures

Figures reproduced from arXiv: 2607.22205 by the authors.

Figure 1
Figure 1. Motivation and paradigm of the proposed FBA. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 3
Figure 3. RS semantic anchoring establishes broad overhead [PITH_FULL_IMAGE:figures/full_fig_p002_3.png] view at source ↗
Figure 2
Figure 2. Progressive data curation of the CPRS dataset for the staged route. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figures from the paper (2 more)
Figure 3
Figure 3. Figure 3: Convergent staged post-training route: S1 RS semantic anchoring, S2 domain-bridge convergence, and S3 evidence [PITH_FULL_IMAGE:figures/full_fig_p004_3.png]
Figure 4
Figure 4. Figure 4: The representation proximity between the Harbor [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

97 extracted references · 5 linked inside Pith

  1. [3]

    Proceedings of the 26th Annual International Conference on Machine Learning , pages =

    Bengio, Yoshua and Louradour, Jerome and Collobert, Ronan and Weston, Jason , title =. Proceedings of the 26th Annual International Conference on Machine Learning , pages =. 2009 , url =

    2009

  2. [4]

    Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing , pages =

    Chan, David and Petryk, Suzanne and Gonzalez, Joseph and Darrell, Trevor and Canny, John , title =. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing , pages =. 2023 , doi =

    2023

  3. [5]

    IEEE Transactions on Geoscience and Remote Sensing , volume =

    Cheng, Qimin and Huang, Haiyan and Xu, Yuan and Zhou, Yuzhuo and Li, Huanying and Wang, Zhongyuan , title =. IEEE Transactions on Geoscience and Remote Sensing , volume =. 2022 , doi =

    2022

  4. [6]

    ISPRS Journal of Photogrammetry and Remote Sensing , volume =

    Ge, Junyao and Zhang, Xu and Zheng, Yang and Guo, Kaitai and Liang, Jimin , title =. ISPRS Journal of Photogrammetry and Remote Sensing , volume =. 2025 , doi =

    2025

  5. [7]

    , title =

    Gururangan, Suchin and Marasovic, Ana and Swayamdipta, Swabha and Lo, Kyle and Beltagy, Iz and Downey, Doug and Smith, Noah A. , title =. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics , pages =. 2020 , doi =

    2020

  6. [8]

    and Shen, Yelong and Wallis, Phillip and Allen-Zhu, Zeyuan and Li, Yuanzhi and Wang, Shean and Wang, Lu and Chen, Weizhu , title =

    Hu, Edward J. and Shen, Yelong and Wallis, Phillip and Allen-Zhu, Zeyuan and Li, Yuanzhi and Wang, Shean and Wang, Lu and Chen, Weizhu , title =. International Conference on Learning Representations , year =

  7. [9]

    ISPRS Journal of Photogrammetry and Remote Sensing , volume =

    Hu, Yuan and Yuan, Jianlong and Wen, Congcong and Lu, Xiaonan and Liu, Yu and Li, Xiang , title =. ISPRS Journal of Photogrammetry and Remote Sensing , volume =. 2025 , doi =

    2025

  8. [10]

    and Milan, Kieran and Quan, John and Ramalho, Tiago and Grabska-Barwinska, Agnieszka and others , title =

    Kirkpatrick, James and Pascanu, Razvan and Rabinowitz, Neil and Veness, Joel and Desjardins, Guillaume and Rusu, Andrei A. and Milan, Kieran and Quan, John and Ramalho, Tiago and Grabska-Barwinska, Agnieszka and others , title =. Proceedings of the National Academy of Sciences , volume =. 2017 , doi =

    2017

Show all 97 references
  1. [11]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =

    Kuckreja, Kartik and Danish, Muhammad Sohail and Naseer, Muzammal and Das, Abhijit and Khan, Salman and Khan, Fahad Shahbaz , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =. 2024 , doi =

    2024

  2. [12]

    Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing , pages =

    Li, Yifan and Du, Yifan and Zhou, Kun and Wang, Jinpeng and Zhao, Xin and Wen, Ji-Rong , title =. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing , pages =. 2023 , doi =

    2023

  3. [13]

    Advances in Neural Information Processing Systems 37 , pages =

    Li, Xiang and Ding, Jian and Elhoseiny, Mohamed , title =. Advances in Neural Information Processing Systems 37 , pages =. 2024 , doi =

    2024

  4. [14]

    ISPRS Journal of Photogrammetry and Remote Sensing , volume =

    Li, Zhenshi and Muhtar, Dilxat and Gu, Feng and He, Yanglangxing and Zhang, Xueliang and Xiao, Pengfeng and He, Guangjun and Zhu, Xiaoxiang , title =. ISPRS Journal of Photogrammetry and Remote Sensing , volume =. 2025 , doi =

    2025

  5. [16]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =

    Liu, Haotian and Li, Chunyuan and Li, Yuheng and Lee, Yong Jae , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =. 2024 , doi =

    2024

  6. [17]

    IEEE Transactions on Geoscience and Remote Sensing , volume =

    Liu, Fan and Chen, Delong and Guan, Zhangqingyun and Zhou, Xiaocong and Zhu, Jiale and Ye, Qiaolin and Fu, Liyong and Zhou, Jun , title =. IEEE Transactions on Geoscience and Remote Sensing , volume =. 2024 , doi =

    2024

  7. [18]

    IEEE Transactions on Geoscience and Remote Sensing , volume =

    Lobry, Sylvain and Marcos, Diego and Murray, Jesse and Tuia, Devis , title =. IEEE Transactions on Geoscience and Remote Sensing , volume =. 2020 , doi =

    2020

  8. [19]

    IEEE Transactions on Geoscience and Remote Sensing , volume =

    Lu, Xiaoqiang and Wang, Binqiang and Zheng, Xiangtao and Li, Xuelong , title =. IEEE Transactions on Geoscience and Remote Sensing , volume =. 2018 , doi =

    2018

  9. [21]

    Proceedings of the 38th International Conference on Machine Learning , pages =

    Radford, Alec and Kim, Jong Wook and Hallacy, Chris and Ramesh, Aditya and Goh, Gabriel and Agarwal, Sandhini and Sastry, Girish and Askell, Amanda and Mishkin, Pamela and Clark, Jack and Krueger, Gretchen and Sutskever, Ilya , title =. Proceedings of the 38th International Co...

    2021

  10. [22]

    and Wayne, Gregory , title =

    Rolnick, David and Ahuja, Arun and Schwarz, Jonathan and Lillicrap, Timothy P. and Wayne, Gregory , title =. Advances in Neural Information Processing Systems 32 , year =

  11. [23]

    , title =

    Soni, Sagar and Dudhane, Akshay and Debary, Hiyam and Fiaz, Mustansar and Munir, Muhammad Akhtar and Danish, Muhammad Sohail and Fraccaro, Paolo and Watson, Campbell D. , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =. 2025 , doi =

    2025

  12. [24]

    Information Fusion , volume =

    Vivone, Gemine , title =. Information Fusion , volume =. 2023 , doi =

    2023

  13. [25]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume =

    Wang, Zhecheng and Prabha, Rajanie and Huang, Tianyuan and Wu, Jiajun and Rajagopal, Ram , title =. Proceedings of the AAAI Conference on Artificial Intelligence , volume =. 2024 , doi =

    2024

  14. [26]

    IEEE Transactions on Geoscience and Remote Sensing , volume =

    Yuan, Zhiqiang and Zhang, Wenkai and Fu, Kun and Li, Xuan and Deng, Chubo and Wang, Hongqi and Sun, Xian , title =. IEEE Transactions on Geoscience and Remote Sensing , volume =. 2022 , doi =

    2022

  15. [27]

    Earth System Science Data , volume =

    Yuan, Zhenghang and Xiong, Zhitong and Mou, Lichao and Zhu, Xiao Xiang , title =. Earth System Science Data , volume =. 2025 , doi =

    2025

  16. [28]

    ISPRS Journal of Photogrammetry and Remote Sensing , volume =

    Zhan, Yang and Xiong, Zhitong and Yuan, Yuan , title =. ISPRS Journal of Photogrammetry and Remote Sensing , volume =. 2025 , doi =

    2025

  17. [29]

    IEEE Transactions on Geoscience and Remote Sensing , volume =

    Zhang, Wei and Cai, Miaoxin and Zhang, Tong and Zhuang, Yin and Mao, Xuerui , title =. IEEE Transactions on Geoscience and Remote Sensing , volume =. 2024 , doi =

    2024

  18. [30]

    IEEE Transactions on Geoscience and Remote Sensing , volume =

    Zhang, Zilun and Zhao, Tiancheng and Guo, Yulong and Yin, Jianwei , title =. IEEE Transactions on Geoscience and Remote Sensing , volume =. 2024 , doi =

    2024

  19. [31]

    and others , title =

    Zheng, Lianmin and Chiang, Wei-Lin and Sheng, Ying and Zhuang, Siyuan and Wu, Zhanghao and Zhuang, Yonghao and Lin, Zi and Li, Zhuohan and Li, Dacheng and Xing, Eric P. and others , title =. Advances in Neural Information Processing Systems 36 , pages =. 2023 , doi =

    2023

  20. [32]

    IEEE Geoscience and Remote Sensing Magazine , volume =

    Zhu, Xiao Xiang and Montazeri, Sina and Ali, Mohsin and Hua, Yuansheng and Wang, Yuanyuan and Mou, Lichao and Shi, Yilei and Xu, Feng and Bamler, Richard , title =. IEEE Geoscience and Remote Sensing Magazine , volume =. 2021 , doi =

    2021

  21. [33]

    Advances in Neural Information Processing Systems 35 , pages =

    Alayrac, Jean-Baptiste and Donahue, Jeff and Luc, Pauline and Miech, Antoine and Barr, Iain and Hasson, Yana and Lenc, Karel and Mensch, Arthur and Millican, Katherine and Reynolds, Malcolm and others , title =. Advances in Neural Information Processing Systems 35 , pages =. 2...

    2022

  22. [34]

    Advances in Neural Information Processing Systems 36 , pages =

    Dai, Wenliang and Li, Junnan and Li, Dongxu and Tiong, Anthony Meng Huat and Zhao, Junqi and Wang, Weisheng and Li, Boyang and Fung, Pascale and Hoi, Steven , title =. Advances in Neural Information Processing Systems 36 , pages =. 2023 , url =

    2023

  23. [35]

    IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing , volume =

    Huang, Lanqing and Liu, Bin and Li, Boying and Guo, Weiwei and Yu, Wenhao and Zhang, Zenghui and Yu, Wenxian , title =. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing , volume =. 2018 , doi =

    2018

  24. [36]

    Proceedings of the 40th International Conference on Machine Learning , pages =

    Li, Junnan and Li, Dongxu and Savarese, Silvio and Hoi, Steven , title =. Proceedings of the 40th International Conference on Machine Learning , pages =. 2023 , url =

    2023

  25. [37]

    ISPRS Journal of Photogrammetry and Remote Sensing , volume =

    Li, Ke and Wan, Gang and Cheng, Gong and Meng, Liqiu and Han, Junwei , title =. ISPRS Journal of Photogrammetry and Remote Sensing , volume =. 2020 , doi =

    2020

  26. [38]

    ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences , volume =

    Schmitt, Michael and Hughes, Lloyd Haydn and Qiu, Chunping and Zhu, Xiao Xiang , title =. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences , volume =. 2019 , doi =

    2019

  27. [39]

    2019 IEEE International Geoscience and Remote Sensing Symposium , pages =

    Sumbul, Gencer and Charfuelan, Marcela and Demir, Begum and Markl, Volker , title =. 2019 IEEE International Geoscience and Remote Sensing Symposium , pages =. 2019 , doi =

    2019

  28. [40]

    and Khashabi, Daniel and Hajishirzi, Hannaneh , title =

    Wang, Yizhong and Kordi, Yeganeh and Mishra, Swaroop and Liu, Alisa and Smith, Noah A. and Khashabi, Daniel and Hajishirzi, Hannaneh , title =. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics , pages =. 2023 , doi =

    2023

  29. [41]

    and Guu, Kelvin and Yu, Adams Wei and Lester, Brian and Du, Nan and Dai, Andrew M

    Wei, Jason and Bosma, Maarten and Zhao, Vincent Y. and Guu, Kelvin and Yu, Adams Wei and Lester, Brian and Du, Nan and Dai, Andrew M. and Le, Quoc V. , title =. International Conference on Learning Representations , year =

  30. [42]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =

    Xia, Gui-Song and Bai, Xiang and Ding, Jian and Zhu, Zhen and Belongie, Serge and Luo, Jiebo and Datcu, Mihai and Pelillo, Marcello and Zhang, Liangpei , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =. 2018 , doi =

    2018

  31. [43]

    Computer Vision -- ECCV 2024 , pages =

    Muhtar, Dilxat and Li, Zhenshi and Gu, Feng and Zhang, Xueliang and Xiao, Pengfeng , title =. Computer Vision -- ECCV 2024 , pages =. 2024 , publisher =. doi:10.1007/978-3-031-72904-1_26 , url =

    2024 doi

  32. [44]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume =

    Pang, Chao and Weng, Xingxing and Wu, Jiang and Li, Jiayu and Liu, Yi and Sun, Jiaxing and Li, Weijia and Wang, Shuai and Feng, Litong and Xia, Gui-Song and He, Conghui , title =. Proceedings of the AAAI Conference on Artificial Intelligence , volume =. 2025 , doi =

    2025

  33. [45]

    IEEE Geoscience and Remote Sensing Magazine , volume =

    Li, Xiang and Wen, Congcong and Hu, Yuan and Yuan, Zhenghang and Zhu, Xiao Xiang , title =. IEEE Geoscience and Remote Sensing Magazine , volume =. 2024 , doi =

    2024

  34. [46]

    2025 , month = dec, howpublished =

    2025

  35. [47]

    Journal of Machine Learning Research , volume =

    van der Maaten, Laurens and Hinton, Geoffrey , title =. Journal of Machine Learning Research , volume =. 2008 , url =

    2008

  36. [48]

    IEEE Access , volume =

    Wei, Shunjun and Zeng, Xiangfeng and Qu, Qizhe and Wang, Mou and Su, Hao and Shi, Jun , title =. IEEE Access , volume =. 2020 , doi =

    2020

  37. [50]

    International Conference on Learning Representations , year =

    Zhang, Yi-Fan and Zhang, Huanyu and Tian, Haochen and Fu, Chaoyou and Zhang, Shuangqing and Wu, Junfei and Li, Feng and Wang, Kun and Wen, Qingsong and Zhang, Zhang and Wang, Liang and Jin, Rong and Tan, Tieniu , title =. International Conference on Learning Representations , year =

  38. [51]

    IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing , volume =

    Zhang, Zhengning and Zhang, Licheng and Wang, Yuxuan and Feng, Pengming and He, Ran , title =. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing , volume =. 2021 , doi =

    2021

  39. [52]

    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. In Advances in Neural Information Processing Systems 35, 23716--23736

    2022

  40. [53]

    Bai, S.; Cai, Y.; Chen, R.; Chen, K.; Chen, X.; Cheng, Z.; Deng, L.; Ding, W.; Gao, C.; Ge, C.; et al. 2025 a . Qwen3-VL Technical Report. arXiv preprint arXiv:2511.21631

    2025 arXiv

  41. [54]

    Bai, S.; Chen, K.; Liu, X.; Wang, J.; Ge, W.; Song, S.; Dang, K.; Wang, P.; Wang, S.; Tang, J.; et al. 2025 b . Qwen2.5-VL Technical Report. arXiv preprint arXiv:2502.13923

    2025 arXiv

  42. [55]

    Bengio, Y.; Louradour, J.; Collobert, R.; and Weston, J. 2009. Curriculum Learning. In Proceedings of the 26th Annual International Conference on Machine Learning, 41--48

    2009

  43. [56]

    ByteDance Seed Team . 2025. Seed1.8 : A Generalized Agentic Model. Official model release. Accessed: 2026-07-14

    2025

  44. [57]

    Chan, D.; Petryk, S.; Gonzalez, J.; Darrell, T.; and Canny, J. 2023. CLAIR : Evaluating Image Captions with Large Language Models. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, 13638--13646

    2023

  45. [58]

    Cheng, Q.; Huang, H.; Xu, Y.; Zhou, Y.; Li, H.; and Wang, Z. 2022. NWPU-Captions Dataset and MLCA-Net for Remote Sensing Image Captioning. IEEE Transactions on Geoscience and Remote Sensing, 60: 1--19

    2022

  46. [59]

    Dai, W.; Li, J.; Li, D.; Tiong, A. M. H.; Zhao, J.; Wang, W.; Li, B.; Fung, P.; and Hoi, S. 2023. InstructBLIP : Towards General-Purpose Vision-Language Models with Instruction Tuning. In Advances in Neural Information Processing Systems 36, 49250--49267

    2023

  47. [60]

    Ge, J.; Zhang, X.; Zheng, Y.; Guo, K.; and Liang, J. 2025. RSTeller : Scaling up Visual Language Modeling in Remote Sensing with Rich Linguistic Semantics from Openly Available Data and Large Language Models. ISPRS Journal of Photogrammetry and Remote Sensing, 226: 146--163

    2025

  48. [61]

    Gururangan, S.; Marasovic, A.; Swayamdipta, S.; Lo, K.; Beltagy, I.; Downey, D.; and Smith, N. A. 2020. Don't Stop Pretraining: Adapt Language Models to Domains and Tasks. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, 8342--8360

    2020

  49. [62]

    J.; Shen, Y.; Wallis, P.; Allen-Zhu, Z.; Li, Y.; Wang, S.; Wang, L.; and Chen, W

    Hu, E. J.; Shen, Y.; Wallis, P.; Allen-Zhu, Z.; Li, Y.; Wang, S.; Wang, L.; and Chen, W. 2022. LoRA : Low-Rank Adaptation of Large Language Models. In International Conference on Learning Representations

    2022

  50. [63]

    Hu, Y.; Yuan, J.; Wen, C.; Lu, X.; Liu, Y.; and Li, X. 2025. RSGPT : A Remote Sensing Vision Language Model and Benchmark. ISPRS Journal of Photogrammetry and Remote Sensing, 224: 272--286

    2025

  51. [64]

    Huang, L.; Liu, B.; Li, B.; Guo, W.; Yu, W.; Zhang, Z.; and Yu, W. 2018. OpenSARShip : A Dataset Dedicated to Sentinel-1 Ship Interpretation. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 11(1): 195--208

    2018

  52. [65]

    A.; Milan, K.; Quan, J.; Ramalho, T.; Grabska-Barwinska, A.; et al

    Kirkpatrick, J.; Pascanu, R.; Rabinowitz, N.; Veness, J.; Desjardins, G.; Rusu, A. A.; Milan, K.; Quan, J.; Ramalho, T.; Grabska-Barwinska, A.; et al. 2017. Overcoming Catastrophic Forgetting in Neural Networks. Proceedings of the National Academy of Sciences, 114(13): 3521--3526

    2017

  53. [66]

    S.; Naseer, M.; Das, A.; Khan, S.; and Khan, F

    Kuckreja, K.; Danish, M. S.; Naseer, M.; Das, A.; Khan, S.; and Khan, F. S. 2024. GeoChat : Grounded Large Vision-Language Model for Remote Sensing. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 27831--27840

    2024

  54. [67]

    Li, J.; Li, D.; Savarese, S.; and Hoi, S. 2023 a . BLIP-2 : Bootstrapping Language-Image Pre-Training with Frozen Image Encoders and Large Language Models. In Proceedings of the 40th International Conference on Machine Learning, 19730--19742

    2023

  55. [68]

    Li, K.; Wan, G.; Cheng, G.; Meng, L.; and Han, J. 2020. Object Detection in Optical Remote Sensing Images: A Survey and A New Benchmark. ISPRS Journal of Photogrammetry and Remote Sensing, 159: 296--307

    2020

  56. [69]

    Li, X.; Ding, J.; and Elhoseiny, M. 2024. VRSBench : A Versatile Vision-Language Benchmark Dataset for Remote Sensing Image Understanding. In Advances in Neural Information Processing Systems 37, 3229--3242

    2024

  57. [70]

    Li, X.; Wen, C.; Hu, Y.; Yuan, Z.; and Zhu, X. X. 2024. Vision-Language Models in Remote Sensing: Current Progress and Future Trends. IEEE Geoscience and Remote Sensing Magazine, 12(2): 32--66

    2024

  58. [71]

    Li, Y.; Du, Y.; Zhou, K.; Wang, J.; Zhao, X.; and Wen, J.-R. 2023 b . Evaluating Object Hallucination in Large Vision-Language Models. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, 292--305

    2023

  59. [72]

    Li, Z.; Muhtar, D.; Gu, F.; He, Y.; Zhang, X.; Xiao, P.; He, G.; and Zhu, X. 2025. LHRS-Bot-Nova : Improved Multimodal Large Language Model for Remote Sensing Vision-Language Interpretation. ISPRS Journal of Photogrammetry and Remote Sensing, 227: 539--550

    2025

  60. [73]

    Liu, F.; Chen, D.; Guan, Z.; Zhou, X.; Zhu, J.; Ye, Q.; Fu, L.; and Zhou, J. 2024 a . RemoteCLIP : A Vision Language Foundation Model for Remote Sensing. IEEE Transactions on Geoscience and Remote Sensing, 62: 1--16

    2024

  61. [74]

    Liu, H.; Li, C.; Li, Y.; and Lee, Y. J. 2024 b . Improved Baselines with Visual Instruction Tuning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 26286--26296

    2024

  62. [75]

    Liu, H.; Li, C.; Wu, Q.; and Lee, Y. J. 2023. Visual Instruction Tuning. arXiv preprint arXiv:2304.08485

    2023 arXiv

  63. [76]

    Lobry, S.; Marcos, D.; Murray, J.; and Tuia, D. 2020. RSVQA : Visual Question Answering for Remote Sensing Data. IEEE Transactions on Geoscience and Remote Sensing, 58(12): 8555--8566

    2020

  64. [77]

    Lu, X.; Wang, B.; Zheng, X.; and Li, X. 2018. Exploring Models and Data for Remote Sensing Image Caption Generation. IEEE Transactions on Geoscience and Remote Sensing, 56(4): 2183--2195

    2018

  65. [78]

    Luo, J.; Pang, Z.; Zhang, Y.; Wang, T.; Wang, L.; Dang, B.; Lao, J.; Wang, J.; Chen, J.; Tan, Y.; and Li, Y. 2024. SkySenseGPT : A Fine-Grained Instruction Tuning Dataset and Model for Remote Sensing Vision-Language Understanding. arXiv preprint arXiv:2406.10100

    2024 arXiv

  66. [79]

    Muhtar, D.; Li, Z.; Gu, F.; Zhang, X.; and Xiao, P. 2024. LHRS-Bot : Empowering Remote Sensing with VGI -Enhanced Large Multimodal Language Model. In Computer Vision -- ECCV 2024, 440--457. Springer Nature Switzerland

    2024

  67. [80]

    Pang, C.; Weng, X.; Wu, J.; Li, J.; Liu, Y.; Sun, J.; Li, W.; Wang, S.; Feng, L.; Xia, G.-S.; and He, C. 2025. VHM : Versatile and Honest Vision Language Model for Remote Sensing Image Analysis. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 39, 6381--6388

    2025

  68. [81]

    W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; Krueger, G.; and Sutskever, I

    Radford, A.; Kim, J. W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; Krueger, G.; and Sutskever, I. 2021. Learning Transferable Visual Models From Natural Language Supervision. In Proceedings of the 38th International Conferen...

    2021

  69. [82]

    P.; and Wayne, G

    Rolnick, D.; Ahuja, A.; Schwarz, J.; Lillicrap, T. P.; and Wayne, G. 2019. Experience Replay for Continual Learning. In Advances in Neural Information Processing Systems 32

    2019

  70. [83]

    H.; Qiu, C.; and Zhu, X

    Schmitt, M.; Hughes, L. H.; Qiu, C.; and Zhu, X. X. 2019. SEN12MS : A Curated Dataset of Georeferenced Multi-Spectral Sentinel-1/2 Imagery for Deep Learning and Data Fusion. In ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, volume IV-2/W7,...

    2019

  71. [84]

    A.; Danish, M

    Soni, S.; Dudhane, A.; Debary, H.; Fiaz, M.; Munir, M. A.; Danish, M. S.; Fraccaro, P.; and Watson, C. D. 2025. EarthDial : Turning Multi-Sensory Earth Observations to Interactive Dialogues. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, ...

    2025

  72. [85]

    Sumbul, G.; Charfuelan, M.; Demir, B.; and Markl, V. 2019. BigEarthNet : A Large-Scale Benchmark Archive for Remote Sensing Image Understanding. In 2019 IEEE International Geoscience and Remote Sensing Symposium, 5901--5904

    2019

  73. [86]

    van der Maaten, L.; and Hinton, G. 2008. Visualizing Data Using t-SNE . Journal of Machine Learning Research, 9: 2579--2605

    2008

  74. [87]

    Vivone, G. 2023. Multispectral and Hyperspectral Image Fusion in Remote Sensing: A Survey. Information Fusion, 89: 405--417

    2023

  75. [88]

    Wang, S.; Zou, X.; Li, K.; Xing, J.; and Tao, P. 2023 a . PanBench : Towards High-Resolution and High-Performance Pansharpening. arXiv preprint arXiv:2311.12083

    2023 arXiv

  76. [89]

    A.; Khashabi, D.; and Hajishirzi, H

    Wang, Y.; Kordi, Y.; Mishra, S.; Liu, A.; Smith, N. A.; Khashabi, D.; and Hajishirzi, H. 2023 b . Self-Instruct: Aligning Language Models with Self-Generated Instructions. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics, 13484--13508

    2023

  77. [90]

    Wang, Z.; Prabha, R.; Huang, T.; Wu, J.; and Rajagopal, R. 2024. SkyScript : A Large and Semantically Diverse Vision-Language Dataset for Remote Sensing. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, 5805--5813

    2024

  78. [91]

    Y.; Guu, K.; Yu, A

    Wei, J.; Bosma, M.; Zhao, V. Y.; Guu, K.; Yu, A. W.; Lester, B.; Du, N.; Dai, A. M.; and Le, Q. V. 2022. Finetuned Language Models Are Zero-Shot Learners. In International Conference on Learning Representations

    2022

  79. [92]

    Wei, S.; Zeng, X.; Qu, Q.; Wang, M.; Su, H.; and Shi, J. 2020. HRSID : A High-Resolution SAR Images Dataset for Ship Detection and Instance Segmentation. IEEE Access, 8: 120234--120254

    2020

  80. [93]

    Xia, G.-S.; Bai, X.; Ding, J.; Zhu, Z.; Belongie, S.; Luo, J.; Datcu, M.; Pelillo, M.; and Zhang, L. 2018. DOTA : A Large-Scale Dataset for Object Detection in Aerial Images. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 3974--3983

    2018

  81. [94]

    Yuan, Z.; Xiong, Z.; Mou, L.; and Zhu, X. X. 2025. ChatEarthNet : A Global-Scale Image-Text Dataset Empowering Vision-Language Geo-Foundation Models. Earth System Science Data, 17(3): 1245--1263

    2025

  82. [95]

    Yuan, Z.; Zhang, W.; Fu, K.; Li, X.; Deng, C.; Wang, H.; and Sun, X. 2022. Exploring a Fine-Grained Multiscale Method for Cross-Modal Remote Sensing Image Retrieval. IEEE Transactions on Geoscience and Remote Sensing, 60: 1--19

    2022

  83. [96]

    Zhan, Y.; Xiong, Z.; and Yuan, Y. 2025. SkyEyeGPT : Unifying Remote Sensing Vision-Language Tasks via Instruction Tuning with Large Language Model. ISPRS Journal of Photogrammetry and Remote Sensing, 221: 64--77

    2025

  84. [97]

    Zhang, W.; Cai, M.; Zhang, T.; Zhuang, Y.; and Mao, X. 2024 a . EarthGPT : A Universal Multimodal Large Language Model for Multisensor Image Comprehension in Remote Sensing Domain. IEEE Transactions on Geoscience and Remote Sensing, 62: 1--20

    2024

  85. [98]

    Zhang, Y.-F.; Zhang, H.; Tian, H.; Fu, C.; Zhang, S.; Wu, J.; Li, F.; Wang, K.; Wen, Q.; Zhang, Z.; Wang, L.; Jin, R.; and Tan, T. 2025. MME-RealWorld : Could Your Multimodal LLM Challenge High-Resolution Real-World Scenarios that are Difficult for Humans? In International Con...

    2025

  86. [99]

    Zhang, Z.; Zhang, L.; Wang, Y.; Feng, P.; and He, R. 2021. ShipRSImageNet : A Large-Scale Fine-Grained Dataset for Ship Detection in High-Resolution Optical Remote Sensing Images. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 14: 8458--8472

    2021

  87. [100]

    Zhang, Z.; Zhao, T.; Guo, Y.; and Yin, J. 2024 b . RS5M and GeoRSCLIP : A Large-Scale Vision-Language Dataset and a Large Vision-Language Model for Remote Sensing. IEEE Transactions on Geoscience and Remote Sensing, 62: 1--23

    2024

  88. [101]

    P.; et al

    Zheng, L.; Chiang, W.-L.; Sheng, Y.; Zhuang, S.; Wu, Z.; Zhuang, Y.; Lin, Z.; Li, Z.; Li, D.; Xing, E. P.; et al. 2023. Judging LLM -as-a-Judge with MT-Bench and Chatbot Arena. In Advances in Neural Information Processing Systems 36, 46595--46623

    2023

  89. [102]

    X.; Montazeri, S.; Ali, M.; Hua, Y.; Wang, Y.; Mou, L.; Shi, Y.; Xu, F.; and Bamler, R

    Zhu, X. X.; Montazeri, S.; Ali, M.; Hua, Y.; Wang, Y.; Mou, L.; Shi, Y.; Xu, F.; and Bamler, R. 2021. Deep Learning Meets SAR : Concepts, Models, Pitfalls, and Perspectives. IEEE Geoscience and Remote Sensing Magazine, 9(4): 143--172

    2021

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.