Forest-Chat: Adapting Vision-Language Agents for Interactive Forest Change Analysis

Ce Zhang; James Brock; Nantheera Anantrasirichai

REVIEW 4 major objections 5 minor 11 references

Forest-Chat: Adapting Vision-Language Agents for Interactive Forest Change Analysis

T0 review · 4 major / 5 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read The paper claims that an LLM-orchestrated agent combining supervised and zero-shot vision modules can provide interactive forest change detection and captioning, and introduces a forest-change dataset to train and evaluate it.

desk verdict Forest-Chat is a credible integration of existing perception modules plus a genuinely new but small forest-change dataset; the headline numbers come from the modules, not the agent, and the dataset split likely leaks overlapping crops into train and test. read the letter →

arxiv 2601.14637 v2 pith:UT6AFGCW submitted 2026-01-21 cs.CV cs.AIcs.CLcs.HC

James Brock , Ce Zhang , Nantheera Anantrasirichai This is my paper

classification cs.CVcs.AIcs.CLcs.HC

keywords forestchangedetectioncaptioningvision-languagemodelLLMagentzero-shotremotesensinginterpretationbi-temporalimageryForest-Changedataset

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 tries to show that conversational AI can make satellite-based forest monitoring accessible. It builds Forest-Chat, an agent that lets a user ask questions about a pair of satellite images and get both a pixel-level map of where forest changed and a sentence describing that change. The agent pairs a supervised vision-language model for change detection and captioning with a zero-shot detector and a multimodal LLM for captioning and refinement. To test this, the authors assemble the Forest-Change dataset of bi-temporal satellite images, change masks, and captions, and report that the supervised configuration reaches 67.1% mean IoU and 40.2% BLEU-4 on it, with zero-shot performance reaching 60.15% and 34.0%.

What carries the argument

The core mechanism is the agent's orchestration architecture: a supervised dual-branch model that shares a bi-temporal feature extractor between a change-detection decoder and a change-captioning decoder, a training-free zero-shot detector that matches segmentation proposals across the two images in a latent space, and a multimodal LLM that generates or refines change captions. An LLM controller interprets user queries and invokes these modules as tools. The central object doing the work is the shared bi-temporal representation: a stack of interaction layers that fuse features from the two images and feed both pixel-level masks and textual descriptions.

What would settle it

A site-disjoint split of Forest-Change — training on crops from some sites and testing on crops from held-out sites — would settle whether the mIoU and BLEU-4 scores hold; if scores drop substantially (e.g., mIoU falling to near the no-change prior), the generalization claim fails.

Watch

Extended reading notes

Core claim

On its own terms, the central discovery is that a joint model trained for both change detection and change captioning can transfer from urban to forest scenes, and that zero-shot change captioning by a multimodal LLM is viable only when domain-specific style guidance or refinement is supplied. The paper reports supervised mean IoU and BLEU-4 of 67.10% and 40.17% on Forest-Change and 88.13% and 34.41% on a tree-focused subset of an urban benchmark; zero-shot numbers are 60.15% and 34.00% and 47.32% and 18.23% respectively. It also shows that refinement of supervised captions by an LLM improves n-gram and semantic metrics on the forest dataset, while on the urban subset it improves semantic si

Load-bearing premise

The Forest-Change dataset is built from a small number of unique geographic sites and is split randomly over heavily overlapping crops, so the reported accuracies may overstate how well the models will perform on genuinely new forest regions.

Editorial extensions

If this is right

Forest change analysis can be conducted interactively in natural language, with a single system returning both change masks and captions.
A supervised dual-task model trained on urban change data can generalize to forest change captioning, with stronger detection scores on urban scenes than on forest scenes.
Zero-shot multimodal-LLM captioning of forest change requires style-guided prompting or refinement to reach useful caption quality; general prompting alone nearly fails.
Cross-domain transfer experiments show that zero-shot supervised models collapse across a 30m-to-sub-meter resolution gap, while modest target-domain supervision restores performance.
Caption refinement is a targeted post-processing tool that injects geographic context into supervised captions, though gains depend on the reference caption style.

Reading between the lines

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

A site-disjoint evaluation of Forest-Change (splitting by geographic site rather than randomly over crops) would test whether the reported scores reflect true generalization to new forest regions; the paper's own discussion notes limited unique sites.
The refinement stage suggests a general recipe for injecting domain knowledge into vision-language captioning: use a strong LLM to enrich supervised outputs, then evaluate with semantic metrics alongside lexical ones.
The zero-shot detection module's sensitivity to atmospheric and seasonal artifacts suggests that pairing it with a lightweight supervised calibrator could yield a practical semi-supervised monitoring pipeline.
The agent's modular design implies that future forest-monitoring tasks (tree counting, disturbance typing, biomass regression) could be added as new tools without retraining the whole system.

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.

Desk Editor's Note

read the letter

You should know this before reading: Forest-Chat is an honest empirical integration of MCI (supervised CDC), AnyChange (zero-shot change detection), and GPT-4o (captioning/refinement) under an LLM orchestrator. The main contributions are the new Forest-Change dataset and a careful evaluation of zero-shot captioning with style-guided prompts. But the headline accuracy numbers in the abstract come from the individual perception models, not from an end-to-end run of the agent, and there is a real risk that the Forest-Change random split leaks near-duplicate crops from the same small set of sites into train and test.

What is actually new: Forest-Change is, to my knowledge, the first dataset pairing forest change masks with captions, and the code and data are promised publicly. The zero-shot captioning analysis is useful: general prompting gives BLEU-4 near zero on Forest-Change, style-guided prompting jumps to 34, and caption refinement gives modest gains on Forest-Change while hurting n-gram metrics on LEVIR-MCI-Trees. That is a nuanced, non-obvious finding. The supervised experiments are thorough: baselines include BiFA, Chg2Cap, Change3D, and a U-Net SiamDiff; they run ablations on MTL loss balancing, gradient conflict methods, and backbone size; they report inference times. The authors are transparent about dataset limitations and the gap between zero-shot and supervised performance.

Three soft spots, in decreasing order of severity. First, the split. The 334 Forest-Change pairs come from a limited number of unique sites, cropped into overlapping 480x480 patches, and then split randomly 80/10/10. The paper itself notes the limited scene diversity. That means near-duplicate crops almost certainly appear in both train and test, so the reported mIoU/BLEU numbers overstate generalization to new forest regions. A site-disjoint split or overlap statistics would fix this, and I would want to see it before trusting the absolute numbers. Second, the zero-shot hyperparameters (thresholds) were selected by Bayesian search over 20 runs without a described validation split. Unless they used the validation set, the zero-shot results are partially test-set-fitted. The improvements over default are modest, but the protocol needs to be explicit. Third, the paper claims "Forest-Chat achieves" these scores, but the scores come from evaluating the MCI model and the AnyChange+GPT-4o pipeline independently. The agent's orchestration is only shown qualitatively. That is fine for a system paper, but the abstract should not imply end-to-end evaluation.

Who this is for: researchers working on remote sensing change detection and captioning, especially anyone thinking about adapting VLM agents to forest monitoring. The dataset, even small, is a useful resource. The paper deserves a serious referee—it is far from a desk reject—but it needs revision on the split and tuning protocol. I would engage with it; the core integration and dataset are worth having in the literature, and the limitations are fixable.

Referee Report

4 major / 5 minor

Summary. The paper introduces Forest-Chat, an LLM-orchestrated agent for interactive forest change analysis that combines a supervised MCI vision-language model (FC-Supervised) with zero-shot perception modules (AnyChange for pixel-level change detection, GPT-4o for change captioning and refinement). The authors also introduce Forest-Change, a new bi-temporal forest change detection and captioning dataset derived from 334 image pairs, along with tree-focused subsets of LEVIR-MCI and JL1-CD. Extensive experiments compare supervised and zero-shot performance on these datasets, study cross-domain transfer, and ablate multi-task training strategies. The headline claim is that Forest-Chat achieves strong mIoU and BLEU-4 scores in both supervised and zero-shot settings, demonstrating the feasibility of interactive LLM-based forest change analysis.

Significance. If the results hold, the paper makes a useful empirical contribution by providing the first forest-specific RSICI dataset and by systematically evaluating the adaptation of a Change-Agent-style VLM system to forest change analysis. The authors are transparent about many limitations, and the public release of data and code (though the link is not explicit) is a strength. The cross-domain study on JL1-CD-Trees and the multi-task ablation are valuable. However, two load-bearing issues—potential train/test leakage in the Forest-Change benchmark and the absence of an end-to-end evaluation of the agent—mean that the central claims, as stated in the abstract, are not yet fully established.

major comments (4)

[§3.2.1, §3.2.2, §6; Table 1] The Forest-Change dataset is constructed from overlapping 480×480 crops originating from a limited number of unique sites (334 pairs from 1500 images), and the 80/10/10 random split (Section 3.2.2) almost certainly places near-duplicate crops from the same source tile into both training and test sets. This makes the reported supervised mIoU of 67.10 and BLEU-4 of 40.17 on Forest-Change an optimistic estimate of generalization to new forest regions. The paper acknowledges the limited unique sites in Section 6, but does not address split-level leakage. A site-disjoint split with overlap statistics is required to support the benchmark's generalization claims.
[§4.2, §5.2–§5.4, Fig. 8; Abstract] The abstract states that 'Forest-Chat achieves' the reported mIoU and BLEU-4 scores, but all quantitative results are produced by the perception modules in isolation—FC-Supervised is the MCI model, and FC-Zero-shot is AnyChange plus GPT-4o with fixed prompts. The LLM orchestration layer is only demonstrated qualitatively in Figure 8. As a result, the core claim that Forest-Chat as an interactive agent achieves these numbers is not directly supported. Either an end-to-end benchmark of the agent's tool selection, reasoning, and response generation should be provided, or the abstract and conclusion should be reworded to attribute the numbers to the perception components.
[§5.3, Table 3] The zero-shot change detection results rely on a Bayesian hyperparameter search of 20 runs per backbone, but the manuscript does not state whether this search was performed on the validation or test split. If the test set was used for hyperparameter selection, the zero-shot mIoU values (e.g., 60.15 on Forest-Change) are optimistically biased. Additionally, the best configuration from Forest-Change is reused for LEVIR-MCI-Trees without a validation-based selection on that dataset. Please clarify the protocol and, if necessary, re-run the search on validation splits only.
[§4.4, Appendix A, Table 4; Abstract] The zero-shot captioning results highlighted in the abstract (BLEU-4 = 34.00 on Forest-Change) are obtained with the 'style-guided' prompt, which includes dataset-specific few-shot examples, allowed vocabulary, and explicit stylistic constraints. This is not 'zero-shot' in the conventional sense of no target-domain examples; it is a form of few-shot prompting. The general-prompt result, which is truly zero-shot, is much lower (BLEU-4 = 0.00 on Forest-Change). The paper should either relabel the style-guided condition as few-shot and report both separately, or justify why the style-guided condition is still considered zero-shot.

minor comments (5)

[§3.2.3, §6] The rule-based caption generation produces highly templated captions, and training/evaluating with BLEU-4 against such references may reward template-matching rather than semantic understanding. The paper acknowledges this limitation, but a human evaluation or additional diversity metrics would strengthen the captioning conclusions.
[Data availability] The text says 'Datasets and code are publicly available here' and later 'ForestChat' without an actual URL. Please provide explicit links.
[References] Some references are incomplete (e.g., 'Ibrahim et al.' has no year or venue; 'Wegler and Kuenzer' is missing page numbers). Please verify all entries.
[Table 1] The table header states that best results are bold and second-best underlined, but the bold/underline formatting is not visible in the manuscript text. Please ensure the formatting is clear in the final version.
[§5.3, Table 2] The inference time for FC-Zero-shot captioning is marked with an asterisk because it includes API overhead. This is fine, but the reader should be reminded that the zero-shot detection times are for a single model forward pass, not the full agent pipeline.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity — the paper's contributions are empirical system/dataset evaluations, not derived predictions.

full rationale

The paper does not contain a derivation chain in which a predicted quantity is shown to follow from a law or first-principles model. All headline numbers (mIoU 67.10/88.13, BLEU-4 40.17/34.41, zero-shot 60.15/34.00) are empirical evaluations of a modular pipeline (MCI, AnyChange, GPT-4o) on dataset splits. The rule-based Forest-Change captions are generated from mask statistics (Sec. 3.2.3), so high BLEU may reflect template matching rather than deep semantic understanding, but the model is trained and then measured against those captions; the captions are not fitted functions of the model output, so this is a benchmark-validity concern rather than a by-construction equivalence. The zero-shot protocol uses a dataset-specific style-guided prompt and a Bayesian hyperparameter search including the change-confidence threshold (Secs. 4.4, 5.3, App. A); this weakens the label 'zero-shot,' but the paper is transparent about these choices. The paper also acknowledges the Forest-Change data's limited unique sites and programmatic captions (Sec. 6). Data-split leakage from overlapping crops (Sec. 3.2.2) is a real external-validity risk, but again is not circularity: the reported numbers are measurements on that split, not predictions forced by construction. There are no load-bearing self-citations: the MCI/Change-Agent and AnyChange bases are cited external works by other author groups, and no uniqueness theorem or ansatz is imported from the present authors' prior work. Therefore the circularity pass finds no circular step.

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

No theoretical entities are invented. The free parameters are almost entirely evaluation choices: AnyChange thresholds, GPT-4o temperature, and rule-based caption binning. The largest epistemic load sits on the Forest-Change ground truth and random-split validity, which the paper itself acknowledges in limited form.

free parameters (4)

AnyChange zero-shot hyperparameter set = change confidence=145 (SAM)/150 (SAM2); stability score=0.93/0.91; area threshold=0.9; object similarity=60/50
Chosen by 20-run Bayesian search in Section 5.3 with no explicit validation split; these thresholds directly set reported zero-shot mIoU values.
GPT-4o captioning temperature = 0.2
Set in Section 4.4 to reduce output variability across all prompting modes; affects all FC-Zero-shot captioning numbers.
Forest-Change rule-based caption bin thresholds = Not reported
Section 3.2.3: percent forest loss is binned into adjective descriptors; the bin edges are not given in the paper, yet they shape the reference captions every captioning model is scored against.
Points-per-side for SAM point prompts = 16
Fixed in Section 5.3 as a practical trade-off; influences mask proposal density and therefore zero-shot segmentation quality.

assumptions (5)

domain assumption Change masks and captions in the source datasets (Hewarathna et al., LEVIR-MCI, JL1-CD) are accurate ground truth for forest change.
All supervised training and evaluation in Sections 5.2-5.5 inherit these labels without independent verification.
ad hoc to paper Forest-Change's rule-based captions are adequate semantic ground truth for training and evaluating captioning.
Section 3.2.3; the majority of Forest-Change captions are generated programmatically from mask statistics, so BLEU/CIDEr scores partly measure reproduction of rule templates rather than semantic quality.
ad hoc to paper Random 80/10/10 splitting of Forest-Change gives a valid generalization estimate despite overlapping crops and few source sites.
Section 3.2.2; not stated as an assumption but required for the reported test numbers to represent generalization. Section 6 admits limited unique sites.
domain assumption SAM's latent-space cosine similarity reliably indicates semantic change for subtle forest disturbance.
AnyChange (Section 4.3) performs training-free bi-temporal matching; the paper does not validate this assumption for fine-grained forest changes beyond the reported thresholds.
standard math Automatic caption metrics and mIoU are valid proxies for change-interpretation quality.
Used in Section 4.6; human evaluation is not performed, so metric validity is assumed.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Forest-Chat: Adapting Vision-Language Agents for Interactive Forest Change Analysis." pith.science (2026). https://pith.science/paper/UT6AFGCW

@misc{pith2026260114637,
  author       = {Pith},
  title        = {Pith review of: Forest-Chat: Adapting Vision-Language Agents for Interactive Forest Change Analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UT6AFGCW}},
  note         = {Machine review of arXiv:2601.14637}
}

read the original abstract

The increasing availability of high-resolution satellite imagery, together with advances in deep learning, creates new opportunities for forest monitoring workflows. Two central challenges in this domain are pixel-level change detection and semantic change interpretation, particularly for complex forest dynamics. While large language models (LLMs) are increasingly adopted for data exploration, their integration with vision-language models (VLMs) for remote sensing image change interpretation (RSICI) remains underexplored, especially beyond urban environments. This paper introduces Forest-Chat, an LLM-driven agent for forest change analysis, enabling natural language querying across multiple RSICI tasks, including change detection and captioning, object counting, deforestation characterisation, and change reasoning. Forest-Chat builds upon a multi-level change interpretation (MCI) vision-language backbone with LLM-based orchestration, incorporating zero-shot change detection via AnyChange and multimodal LLM-based zero-shot change captioning and refinement. To support adaptation and evaluation in forest environments, we introduce the Forest-Change dataset, comprising bi-temporal satellite imagery, pixel-level change masks, and semantic change captions via human annotation and rule-based methods. Forest-Chat achieves mIoU and BLEU-4 scores of 67.10% and 40.17% on Forest-Change, and 88.13% and 34.41% on LEVIR-MCI-Trees, a tree-focused subset of LEVIR-MCI. In a zero-shot capacity, it achieves 60.15% and 34.00% on Forest-Change, and 47.32% and 18.23% on LEVIR-MCI-Trees. Further experiments demonstrate the value of caption refinement for injecting geographic domain knowledge into supervised captions, and the system's limited label domain transfer onto JL1-CD-Trees. These findings demonstrate that interactive, LLM-driven systems can support accessible and interpretable forest change analysis.

Figures

Figures reproduced from arXiv: 2601.14637 by the authors.

**Figure 2.** Summary statistics of change cover in segmentation masks for the [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** Summary statistics of captions for Forest-Change and LEVIR-MCI [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Figures from the paper (5 more)

**Figure 4.** Figure 4: Four examples from the Forest-Change dataset, that each display a [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

**Figure 5.** Figure 5: Overview of the two visual perception models available to the Forest-Chat agent. (a) Highlights the overall structure of the MCI model (b) Introduces a [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

**Figure 6.** Figure 6: Cross-domain transfer performance on JL1-CD-Trees dataset. a) Models pretrained on Forest-Change and fine-tuned on increasing fractions of JL1-CD [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]

**Figure 7.** Figure 7: A selection of qualitative comparison results between the benchmarked models on the Forest-Change test set. Yellow indicates agreement with GT mask, [PITH_FULL_IMAGE:figures/full_fig_p017_7.png]

**Figure 8.** Figure 8: Conversation examples between users and Forest-Chat. Forest-Chat is capable of understanding users and orchestrating relevant tools to answer questions [PITH_FULL_IMAGE:figures/full_fig_p018_8.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

11 extracted references · 6 linked inside Pith

[5]

Advances in Neural Information Processing Sys- tems 36, 68539–68551

Toolformer: Language models can teach themselves to use tools. Advances in Neural Information Processing Sys- tems 36, 68539–68551. Shabbir, A., Munir, M.A., Dudhane, A., Sheikh, M.U., Khan, M.H., Fraccaro, P., Moreno, J.B., Khan, F.S., Khan, S.,
[6]

arXiv preprint arXiv:2505.23752

Thinkgeo: Evaluating tool-augmented agents for re- mote sensing tasks. arXiv preprint arXiv:2505.23752 . Shi, J., Zhang, M., Hou, Y ., Zhi, R., Liu, J., 2024. A multi-task network and two large scale datasets for change detection and captioning in remote sensing images. IEEE Transactions on Geoscience and Remote Sensing . Soni, S., Dudhane, A., Debary, H....

arXiv 2024
[7]

Remote Sensing 17, 162

Advancements in vision–language models for remote sensing: Datasets, capabilities, and enhancement techniques. Remote Sensing 17, 162. Taylor, R., Dudley, N., Stolton, S., Shapiro, A., 2015. Defor- estation fronts: 11 places where most forest loss is projected between 2010 and 2030, in: Proceedings of the XIV World Forestry Congress, Durban, South Africa,...

arXiv 2015
[8]

arXiv preprint arXiv:2406.07089

Rs-agent: Automating remote sensing tasks through intelligent agent. arXiv preprint arXiv:2406.07089 . Yang, B., Wang, X., Xing, Y ., Cheng, C., Jiang, W., Feng, Q.,

arXiv
[9]

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 17, 17052–17065

Modality fusion vision transformer for hyperspectral and lidar data collaborative classification. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 17, 17052–17065. Yang, C., Li, Z., Jiao, H., Gao, Z., Zhang, L., 2025a. Enhancing perception of key changes in remote sensing image change captioning. IEEE Transactions on Image ...

2023
[2009]

Imagenet: A large-scale hierarchical image database, in: 2009 IEEE conference on computer vision and pattern recognition, Ieee. pp. 248–255. Deng, P., Zhou, W., Wu, H., 2025. Changechat: An inter- active model for remote sensing change analysis via multi- modal instruction tuning, in: ICASSP 2025-2025 IEEE In- ternational Conference on Acoustics, Speech a...

2009
[2019]

the scene is the same as before

Bertscore: Evaluating text generation with bert. arXiv preprint arXiv:1904.09675 . Zhang, W., Cai, M., Zhang, T., Zhuang, Y ., Mao, X., 2024b. 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. Zhang, Z., Shen, H., Zhao, T., Chen, ...

arXiv 1904
[2020]

Advances in neural information processing systems 33, 5824–5836

Gradient surgery for multi-task learning. Advances in neural information processing systems 33, 5824–5836. Yu, Y ., Zhuang, Y ., Zhang, J., Meng, Y ., Ratner, A.J., Kr- ishna, R., Shen, J., Zhang, C., 2023. Large language model as attributed training data generator: A tale of diversity and bias. Advances in neural information processing systems 36, 55734–...

2023

Show all 11 references

[2023]

Advances in Neural Information Processing Systems 36, 5539–5568

Openagi: When llm meets domain experts. Advances in Neural Information Processing Systems 36, 5539–5568. Gorelick, N., Hancher, M., Dixon, M., Ilyushchenko, S., Thau, D., Moore, R., 2017. Google earth engine: Planetary-scale geospatial analysis for everyone. Remote sensing of ...

2017 arXiv
[2024]

Remote Sensing 16, 1477

Rs-llava: A large vision-language model for joint cap- tioning and question answering in remote sensing imagery. Remote Sensing 16, 1477. Cai, C., Xu, H., Chen, S., Yang, L., Weng, Y ., Huang, S., Dong, C., Lou, X., 2023. Tree recognition and crown width ex- traction based on ...

2023 arXiv
[2025]

International Journal of Applied Earth Observation and Geoinformation 136, 104282

Deep learning change detection techniques for opti- cal remote sensing imagery: Status, perspectives and chal- lenges. International Journal of Applied Earth Observation and Geoinformation 136, 104282. Qin, Y ., Ding, S., Xie, H., Tan, B., Li, Y ., 2025. A fine-grained vision-...

2025 arXiv

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.

[1] [5]

Advances in Neural Information Processing Sys- tems 36, 68539–68551

Toolformer: Language models can teach themselves to use tools. Advances in Neural Information Processing Sys- tems 36, 68539–68551. Shabbir, A., Munir, M.A., Dudhane, A., Sheikh, M.U., Khan, M.H., Fraccaro, P., Moreno, J.B., Khan, F.S., Khan, S.,

[2] [6]

arXiv preprint arXiv:2505.23752

Thinkgeo: Evaluating tool-augmented agents for re- mote sensing tasks. arXiv preprint arXiv:2505.23752 . Shi, J., Zhang, M., Hou, Y ., Zhi, R., Liu, J., 2024. A multi-task network and two large scale datasets for change detection and captioning in remote sensing images. IEEE Transactions on Geoscience and Remote Sensing . Soni, S., Dudhane, A., Debary, H....

arXiv 2024

[3] [7]

Remote Sensing 17, 162

Advancements in vision–language models for remote sensing: Datasets, capabilities, and enhancement techniques. Remote Sensing 17, 162. Taylor, R., Dudley, N., Stolton, S., Shapiro, A., 2015. Defor- estation fronts: 11 places where most forest loss is projected between 2010 and 2030, in: Proceedings of the XIV World Forestry Congress, Durban, South Africa,...

arXiv 2015

[4] [8]

arXiv preprint arXiv:2406.07089

Rs-agent: Automating remote sensing tasks through intelligent agent. arXiv preprint arXiv:2406.07089 . Yang, B., Wang, X., Xing, Y ., Cheng, C., Jiang, W., Feng, Q.,

arXiv

[5] [9]

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 17, 17052–17065

Modality fusion vision transformer for hyperspectral and lidar data collaborative classification. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 17, 17052–17065. Yang, C., Li, Z., Jiao, H., Gao, Z., Zhang, L., 2025a. Enhancing perception of key changes in remote sensing image change captioning. IEEE Transactions on Image ...

2023

[6] [2009]

Imagenet: A large-scale hierarchical image database, in: 2009 IEEE conference on computer vision and pattern recognition, Ieee. pp. 248–255. Deng, P., Zhou, W., Wu, H., 2025. Changechat: An inter- active model for remote sensing change analysis via multi- modal instruction tuning, in: ICASSP 2025-2025 IEEE In- ternational Conference on Acoustics, Speech a...

2009

[7] [2019]

the scene is the same as before

Bertscore: Evaluating text generation with bert. arXiv preprint arXiv:1904.09675 . Zhang, W., Cai, M., Zhang, T., Zhuang, Y ., Mao, X., 2024b. 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. Zhang, Z., Shen, H., Zhao, T., Chen, ...

arXiv 1904

[8] [2020]

Advances in neural information processing systems 33, 5824–5836

Gradient surgery for multi-task learning. Advances in neural information processing systems 33, 5824–5836. Yu, Y ., Zhuang, Y ., Zhang, J., Meng, Y ., Ratner, A.J., Kr- ishna, R., Shen, J., Zhang, C., 2023. Large language model as attributed training data generator: A tale of diversity and bias. Advances in neural information processing systems 36, 55734–...

2023