Pith. sign in

REVIEW 4 major objections 4 minor 78 references

Single-tool zoom-in saturates on hard satellite-reasoning tasks; the paper's answer is a multi-tool loop—crop, ground, and draw lines—trained with attention-focused reinforcement learning, which beats far larger models on three benchmarks.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 00:53 UTC pith:O4NFWJLN

load-bearing objection GeoMTVR is a genuinely useful dataset and the pilot study is thoughtful, but RTAL's Eq. (3) as written has zero gradient at the start of every update, so the +1.8% headline gain is unexplained. the 4 major comments →

arxiv 2607.25993 v1 pith:O4NFWJLN submitted 2026-07-28 cs.CV

Beyond Zooming: Learning Multi-Tool Visual Reasoning for Ultra-High-Resolution Remote Sensing

Fengxiang Wang , Jiangnan Huang , Mingshuo Chen , Yueying Li , Yang Shi , Junwei Luo , Haoyu Wang , Yansheng Li
show 3 more authors
Jing Zhang Haiyan Zhao Wenjing Yang
This is my paper
classification cs.CV
keywords ultra-high-resolution remote sensingmulti-tool visual reasoningzoom-inobject groundingauxiliary linesreinforcement learningvisual question answeringtool-attention learning
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Ultra-high-resolution satellite images are too large for a model to answer questions from a single global view, and this paper claims that the usual fix—letting the model zoom in—only solves easy cases. Through a pilot study on the XLRS-Bench benchmark, the authors show that zoom-in saturates on tasks requiring global counting, multi-region comparison, path planning, or evidence scattered across the image. To go beyond zoom, they build GeoMTVR, a dataset of 13,000 satellite question-answer samples annotated with interleaved reasoning traces that call three tools: crop-and-zoom, object grounding, and auxiliary-line drawing. They then train GeoLens in two stages—supervised fine-tuning on those traces, followed by a reinforcement-learning step that concentrates gradients on the attention patterns at tool-call tokens. On three benchmarks, GeoLens outperforms direct-reasoning and zoom-only baselines and, with a 7-billion-parameter model, beats closed models two orders of magnitude larger.

Core claim

The paper's central claim is that evidence acquisition in ultra-high-resolution remote sensing is a multi-tool problem, not a resolution problem. The authors report that a single zoom-in tool resolves tasks whose evidence is locally recoverable, such as object classification, regional counting, and attribute recognition, but leaves a 'moving failure set' behind: as training solves easy failures, the remaining hard cases require global coverage, comparison across distant regions, and structured spatial operations that one crop cannot provide. The proposed solution is to let the model also call an object-grounding tool and an auxiliary-line tool, and to train the policy with an attention-level

What carries the argument

The engine of the paper is the combination of GeoMTVR, a supervised dataset of 13,078 ultra-high-resolution VQA samples with interleaved tool-use trajectories, and Reinforced Tool Attention Learning (RTAL). RTAL is the key mechanism: after standard supervised fine-tuning, it extracts the time steps inside tool-call spans, normalizes the model's final-layer attention over the valid causal context to form a 'tool-conditioned causal attention policy,' and applies an advantage-weighted Jensen–Shannon divergence so that high-reward trajectories pull the attention routing at tool tokens toward successful patterns while low-reward trajectories push it away. A time-step mask restricts backpropagatio

Load-bearing premise

The staged annotation pipeline produces correct, non-hallucinated tool-use trajectories; if the coarse crops, generated traces, or box localizations are systematically wrong, supervised fine-tuning teaches the model bad tool behavior and the reinforcement-learning gains are built on that flawed foundation, and the paper does not report a human-quality audit of these traces.

What would settle it

Run GeoLens's RL stage with tool-call tokens replaced by non-functional placeholders (or with returned observations blanked); if accuracy on XLRS-Bench stays near 54.2%, the reported gains come from the reasoning text rather than the tool calls. Alternatively, have human annotators audit a random subset of GeoMTVR trajectories and measure the fraction of tool calls whose outputs actually localize the queried evidence; if that fraction is low, the training signal is contaminated.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • If correct, model scale is not the dominant factor for ultra-high-resolution remote-sensing QA: a 7-billion-parameter model trained on multi-tool reasoning outperforms models with far more parameters on the same benchmarks.
  • Zoom-in should be treated as one tool among several; tasks requiring global coverage, multi-region comparison, or long-range spatial reasoning need grounding and line-drawing operations to gather dispersed evidence.
  • Attention-level reinforcement is a viable credit-assignment mechanism for tool-use policies: concentrating optimization on tool-call spans improves accuracy and yields more decisive, less redundant tool trajectories than token-level RL.
  • Tool-use trajectories can be generated at scale by a staged semi-automatic pipeline that separates coarse question-region cropping from trace generation and object localization, then used to teach active multi-tool reasoning.
  • UHR remote-sensing MLLMs should be designed as active, task-adaptive agents that interleave reasoning with visual tool calls rather than as single-pass image understanders.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: the 'moving failure set' result likely generalizes beyond zoom-in; any domain with a single narrow evidence-gathering operation, such as OCR on documents or depth maps in robot navigation, should show the same pattern of easy wins followed by saturation as tasks require heterogeneous evidence.
  • Editorial inference: RTAL's success implies a testable prediction about internal representations—models trained with RTAL should show measurable differences in attention entropy and cross-step routing at tool tokens compared with token-level RL models, and these differences should correlate with tool-call quality.
  • Editorial inference: because the annotation pipeline is semi-automatic, a cheap quality check is to run the same staged pipeline with and without the coarse-crop stage and measure downstream task accuracy; if traces are sensitive to that stage, dataset noise is likely concentrated in localization, pointing to where further human cleaning would pay off.
  • Editorial inference: explicit tool calls produce inspectable intermediate evidence—crop boxes, object boxes, drawn lines—which could make GeoLens-style systems easier to audit than end-to-end black-box answers, but only if the tool outputs themselves are trustworthy.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper introduces GeoMTVR, a 13K-sample ultra-high-resolution remote-sensing VQA dataset with interleaved multi-tool reasoning traces (crop-and-zoom, grounding, auxiliary-line drawing), and proposes Reinforced Tool Attention Learning (RTAL), an attention-level RL objective intended to focus policy updates on tool-call spans. The resulting GeoLens model (Qwen2.5-VL-7B backbone) is reported to achieve 54.2% on XLRS-Bench, 60.7% on LRS-GRO-eval, and 48.8% on RSHR-Bench, surpassing much larger models including Qwen3-VL-235B. The authors claim that multi-tool visual reasoning with tool-attention RL is the key ingredient, supported by a pilot study showing the limits of single-tool zoom-in and by ablations comparing SFT, GRPO, and RTAL.

Significance. If the reported results hold, the paper makes a substantial empirical contribution: it demonstrates that a 7B model can outperform 235B-scale models on UHR RS benchmarks by combining domain-specific tool-use data with RL, and it releases a first-of-its-kind multi-tool trajectory dataset (GeoMTVR). The pilot-study decomposition into fixed and dynamic failure sets is a useful diagnostic. However, the central algorithmic novelty, RTAL, is not supported by the presented mathematics: Eq. (3) has zero gradient at the stated initialization, so the +1.8-point RTAL-vs-GRPO gain in Table 16 cannot be produced by the objective as written. This is a load-bearing issue that must be addressed before the results can be interpreted.

major comments (4)
  1. [§4.2.1, Eq. (3)] The RTAL objective L_tool_AttnRL = Aτ * (1/|C|) Σ_{t∈C} JSD(pθ_t ∥ pold_t) is inert at the start of each update. Since pold_t is defined as 'the frozen attention policy before the current update', at the beginning of an update we have pθ_t = pold_t for all t, and the JSD and its gradient vanish identically. Thus a positive advantage cannot 'pull' attention toward successful routing and a negative advantage cannot 'push' it away. If multiple inner PPO steps are used, after the first step the term with A>0 acts as a trust-region regularizer pulling back toward the pre-update policy, not as a reinforcement signal; with A<0 it pushes away. The claimed mechanism, and therefore the +1.8-point gain over GRPO in Table 16, is unexplained by Eq. (3). The authors must either correct the objective/reference-policy definition so that the gradient is non-vanishing and has the stated sign, or report th
  2. [Tables 14–16] The headline comparisons lack statistical support. Full-tool SFT vs. zoom-only SFT differs by +0.7 points on average (50.3 vs. 49.6); RTAL vs. GRPO differs by +1.8 points (54.2 vs. 52.4). With 13 sub-tasks, many individual sub-task differences go in opposite directions (e.g., Table 16: GRPO is better on OLUC, RLUC, OCC, OCL; RTAL is better on RP, RCCD, CCR). No error bars, confidence intervals, or multiple-seed runs are reported. The inconsistency between Table 5 (Qwen2.5-VL-7B = 47.4) and Table 15 (same model = 44.9) further undermines the ablations. Please report per-seed variance or bootstrap CIs, and reconcile the base-model numbers.
  3. [§3, Fig. 2] GeoMTVR is built with a staged automatic pipeline: Qwen3-VL-235B coarse cropping, GPT-5.2 trace generation, and GroundingDINO localization. The paper acknowledges that direct MLLM annotation 'often leads to unreliable results', but no human-quality audit is reported for the final GeoMTVR traces. Since SFT and RL are trained on these trajectories, hallucinated or mislocalized tool calls would teach incorrect tool-use behavior. Please provide a human-annotated sample audit (e.g., tool-call correctness, box IoU, line validity, and answer-trace alignment) or another quantitative quality validation.
  4. [§3, Tables 10–11] Data deduplication is described only as 'strict image-level deduplication'. Because GeoMTVR is built from SuperRSVQA and LRS-GRO-train, and evaluation is on XLRS-Bench and LRS-GRO-eval, the procedure for detecting near-duplicates or same-source images is critical. If large scene images are cropped differently or the same source tiles appear in both training and test, the reported SOTA could be inflated. Please clarify the deduplication method, including any perceptual-hash or metadata-based checks, and report the number of removed samples.
minor comments (4)
  1. [§3 / Table 2] The pilot experiment text says 'As shown in Table 1' but the table is numbered Table 2. The arrow notation in Tables 14–16 (e.g., '48.9↓+0.7') is confusing; an upward arrow or explicit 'improvement' would be clearer.
  2. [§3] The text says the average resolution of GeoMTVR is 'about 9K × 9K', but Table 3 reports 6,033×6,017. Please correct the inconsistency.
  3. [Table 6] The column abbreviations (COL, OGD, MRJC, etc.) are not expanded in the caption or main text. Please provide a metric-name table or expand the abbreviations in the caption.
  4. [Appendix A.2.2] The description says the RTAL 'attention regularization' uses tool-action and response masks, but it does not specify how the JSD term in Eq. (3) is computed for the response mask or what λ_attn is set to. Reporting λ_attn and the masking details would help reproducibility.

Circularity Check

0 steps flagged

No equation-level circularity; reported gains are held-out evaluations, and the self-citations are context/baselines rather than load-bearing premises.

full rationale

GeoLens is SFT'd on GeoMTVR (built from SuperRSVQA/LRS-GRO-train with staged Qwen3-VL/GPT-5.2/GroundingDINO annotations) and then RTAL-tuned; it is evaluated on XLRS-Bench, LRS-GRO-eval, and RSHR-Bench. The paper states image-level deduplication and removal of overlap with XLRS-Bench and LRS-GRO-eval (Sec. 3, 'Data source selection'), so the benchmark numbers are not predictions of training labels. RTAL has no free parameters fitted to test answers; its advantage is computed from rollouts. The pilot study on XLRS-Bench is an empirical motivation, not a derivation of GeoLens' SOTA. The self-citations to GeoLLaVA, XLRS-Bench, and GeoEyes supply datasets, benchmarks, and prior zoom-in findings; none is invoked as a uniqueness theorem or as the sole justification for the central result, so the heavy self-citation is a transparency concern rather than circularity. Separate technical issues (e.g., the gradient behavior of Eq. (3) at initialization and absence of a human audit of the annotation pipeline) are correctness/data-quality risks, not reductions of the claimed predictions to their inputs.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 0 invented entities

The central contribution is an ML system, not a physical law; the ledger therefore lists the hand-set coefficients and the domain assumptions about annotation quality, attention-as-policy, and benchmark validity that carry the empirical claims. No new physical or conceptual entities are postulated.

free parameters (4)
  • lambda_attn (attention RL weight) = not stated
    Coefficient coupling L_RL and L_tool_AttnRL in Eq. (4); no value or sensitivity ablation is reported, yet it controls how strongly attention routing is optimized.
  • rollout number n = 8
    RL rollout count per prompt (Table 13); chosen by hand, affects variance and stability of the policy optimization.
  • SFT data mixture = 9,211 SuperVQA + 3,479 LRS-GRO + 388 route_visual
    Composition of GeoMTVR SFT corpus (Table 10) is chosen by availability and source quality; the paper's ablations vary tool annotation, not source mixture.
  • pilot tool budget = up to 5 zoom-in calls
    In the pilot study (Sec. 4.1) the model is allowed up to five zoom-in calls; the 'saturation' conclusion for hard tasks depends on this fixed budget.
axioms (4)
  • domain assumption Qwen3-VL-235B coarse crops, GPT-5.2 tool traces, and GroundingDINO boxes are accurate enough to serve as supervised tool-use trajectories.
    Sec. 3 and Fig. 2 build GeoMTVR semi-automatically from these components; if localization shifts or hallucinated traces are common, SFT teaches invalid tool behavior. The paper notes raw MLLM annotation is unreliable, motivating the staged pipeline, but no human-quality audit is reported.
  • domain assumption Final-layer attention over tool-call spans is a suitable policy for RL optimization.
    Eqs. (2)-(8) define RTAL on normalized final-layer attention. The paper provides no evidence that these attention weights correlate with correct tool-use decisions or that JSD minimization in this subspace improves answer accuracy beyond token-level RL.
  • domain assumption XLRS-Bench, RSHR-Bench, and LRS-GRO-eval are valid, non-contaminated evaluations of UHR RS reasoning.
    The paper asserts image-level deduplication removes overlap with XLRS-Bench and LRS-GRO-eval (Sec. 3), but no code/data is available to verify it, and RSHR-Bench is 'newly introduced' in this paper, making it partially an in-house evaluation.
  • domain assumption Standard MLLM training with a frozen vision encoder can acquire new visual tool policies from interleaved text/image observations.
    GeoLens freezes the vision tower and projector during SFT (Table 12); the claimed multi-tool visual grounding must be learned through the language model's interaction with returned images, which is plausible but not established.

pith-pipeline@v1.3.0-alltime-deepseek · 22082 in / 12978 out tokens · 121186 ms · 2026-08-01T00:53:06.238337+00:00 · methodology

0 comments
read the original abstract

Ultra-high-resolution (UHR) remote-sensing (RS) imagery provides fine-grained Earth-observation evidence over city-scale scenes, but poses a fundamental challenge for multimodal large language models (MLLMs): task-relevant evidence is often sparse, local, and spatially dispersed across extremely large visual contexts. A natural solution is to equip MLLMs with zoom-in tools for active local inspection. However, through a pilot study on XLRS-Bench, we find that zoom-in is only partially effective: it resolves easy and medium-level tasks with locally recoverable evidence, but saturates on hard cases requiring global search, multi-region comparison, path planning, or dispersed-evidence reasoning. Motivated by this finding, we move beyond single-tool zoom-in and introduce GeoMTVR, a large-scale Geospatial Multi-Tool Visual Reasoning dataset built from wide-area satellite imagery. GeoMTVR contains 13K UHR VQA samples with interleaved reasoning trajectories, diverse visual tool calls, and returned visual observations, enabling models to learn question decomposition, tool selection, regional inspection, object-level grounding, auxiliary visual reasoning, and cross-tool evidence integration. Beyond supervised fine-tuning, we propose a tool-attention-focused reinforcement learning algorithm that concentrates optimization on critical tool-use decisions, including when to invoke tools, which tool to select, where to apply it, and how to interpret tool outputs. By combining SFT on GeoMTVR with our RL algorithm, we develop GeoLens, a multi-tool visual reasoning MLLM for UHR RS. Experiments show that GeoLens consistently outperforms direct reasoning and single-tool zoom-in baselines, achieving stronger accuracy, better evidence grounding, and more efficient tool-use trajectories.

Figures

Figures reproduced from arXiv: 2607.25993 by Fengxiang Wang, Haiyan Zhao, Haoyu Wang, Jiangnan Huang, Jing Zhang, Junwei Luo, Mingshuo Chen, Wenjing Yang, Yang Shi, Yansheng Li, Yueying Li.

Figure 1
Figure 1. Figure 1: Overview of GeoLens. GeoLens enables multi-tool visual reasoning for ultra-high￾resolution remote-sensing imagery. fine-grained details, while exhaustive local inspection is computationally inefficient and difficult to learn [8, 1]. A natural solution is to equip MLLMs with visual tools, especially crop-and-zoom, so that they can actively inspect high-resolution local regions during reasoning. Given an ima… view at source ↗
Figure 2
Figure 2. Figure 2: Data Construction Pipeline. Pilot Experiments of Dataset. General-domain datasets have recently provided supervised traces for visual tool use, making it tempting to combine them with high-resolution RS VQA data for training tool-augmented RS MLLMs. We test this strategy by only changing the mix￾ture data in the SFT stage. As shown in Ta￾ble 1, Qwen2.5-VL-7B [49] trained with Su￾perRSVQA [1] alone achieves… view at source ↗
Figure 3
Figure 3. Figure 3: Our dataset covers perception and reasoning dimensions across 13 sub-tasks. Data source selection. Our dataset is built on existing high resolution remote sensing VQA resources. We integrate multi-source imagery and apply unified re-annotation to improve data diversity and training utility. Specifically, we use high-resolution VQA sources including SuperRSVQA [1] and LRS-GRO-train [13]. To ensure data qual… view at source ↗
Figure 4
Figure 4. Figure 4: Quality factor under three failure-set definitions on XLRS-Bench. The gap between fixed and dynamic failure sets reveals the moving failure set effect: as training progresses, easier failures are resolved, while the remaining dynamic failures require more complex evidence acquisition than a single zoom-in tool can provide. We consider three failure sets: the dynamic fail￾ure set Dfail(t), containing sample… view at source ↗
Figure 5
Figure 5. Figure 5: Overview of Reinforced Tool Attention Learning. RTAL focuses reinforcement signals on tool-call spans, encouraging the model to route attention toward task-relevant context and tool observations during multi-tool visual reasoning. 4.2.1 Reinforced Tool Attention Learning In tool-augmented UHR remote-sensing reasoning, the challenge is not only to produce the correct answer, but also to allocate computation… view at source ↗
Figure 6
Figure 6. Figure 6: Visualization of a GeoMTVR sample. The example illustrates the interleaved reasoning trajectory and visual tool use in GeoMTVR. A.5 Potential Societal Impact Potential positive societal impact. GeoLens and GeoMTVR are designed to improve visual evidence acquisition and reasoning over ultra-high-resolution remote-sensing imagery. Such capabilities can support applications where analysts need to inspect larg… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

78 extracted references · 33 linked inside Pith

  1. [1]

    Geollava-8k: Scaling remote-sensing multimodal large language models to 8k resolution

    Fengxiang Wang, Mingshuo Chen, Yueying Li, Di Wang, Haotian Wang, Zonghao Guo, Zefan Wang, Boqi Shan, Long Lan, Yulin Wang, Hongzhen Wang, Wenjing Yang, Bo Du, and Jing Zhang. Geollava-8k: Scaling remote-sensing multimodal large language models to 8k resolution. arXiv preprint arXiv:2505.21375, 2025

    arXiv 2025

  2. [2]

    Xlrs-bench: Could your multimodal llms understand extremely large ultra-high-resolution remote sensing imagery?arXiv preprint arXiv:2503.23771, 2025

    Fengxiang Wang, Hongzhen Wang, Mingshuo Chen, Di Wang, Yulin Wang, Zonghao Guo, Qiang Ma, Long Lan, Wenjing Yang, Jing Zhang, et al. Xlrs-bench: Could your multimodal llms understand extremely large ultra-high-resolution remote sensing imagery?arXiv preprint arXiv:2503.23771, 2025

    Pith/arXiv arXiv 2025

  3. [3]

    A benchmark for ultra-high-resolution remote sensing mllms.arXiv preprint arXiv:2512.17319, 2025

    Yunkai Dang, Meiyi Zhu, Donghao Wang, Yizhuo Zhang, Jiacheng Yang, Qi Fan, Yuekun Yang, Wenbin Li, Feng Miao, and Yang Gao. A benchmark for ultra-high-resolution remote sensing mllms.arXiv preprint arXiv:2512.17319, 2025

    arXiv 2025

  4. [4]

    Geochat: Grounded large vision-language model for remote sensing

    Kartik Kuckreja, Muhammad Sohail Danish, Muzammal Naseer, Abhijit Das, Salman Khan, and Fahad Shahbaz Khan. Geochat: Grounded large vision-language model for remote sensing. 11 InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 27831–27840, 2024

    2024

  5. [5]

    Earthgpt: A universal multi-modal large language model for multi-sensor image comprehension in remote sensing domain.IEEE Transactions on Geoscience and Remote Sensing, 2024

    Wei Zhang, Miaoxin Cai, Tong Zhang, Yin Zhuang, and Xuerui Mao. Earthgpt: A universal multi-modal large language model for multi-sensor image comprehension in remote sensing domain.IEEE Transactions on Geoscience and Remote Sensing, 2024

    2024

  6. [6]

    Rsgpt: A remote sensing vision language model and benchmark.ISPRS Journal of Photogrammetry and Remote Sensing, 224:272–286, 2025

    Yuan Hu, Jianlong Yuan, Congcong Wen, Xiaonan Lu, Yu Liu, and Xiang Li. Rsgpt: A remote sensing vision language model and benchmark.ISPRS Journal of Photogrammetry and Remote Sensing, 224:272–286, 2025

    2025

  7. [7]

    Skyeyegpt: Unifying remote sensing vision- language tasks via instruction tuning with large language model.ISPRS Journal of Photogram- metry and Remote Sensing, 221:64–77, 2025

    Yang Zhan, Zhitong Xiong, and Yuan Yuan. Skyeyegpt: Unifying remote sensing vision- language tasks via instruction tuning with large language model.ISPRS Journal of Photogram- metry and Remote Sensing, 221:64–77, 2025

    2025

  8. [8]

    When large vision-language model meets large remote sensing imagery: Coarse- to-fine text-guided token pruning.arXiv preprint arXiv:2503.07588, 2025

    Junwei Luo, Yingying Zhang, Xue Yang, Kang Wu, Qi Zhu, Lei Liang, Jingdong Chen, and Yansheng Li. When large vision-language model meets large remote sensing imagery: Coarse- to-fine text-guided token pruning.arXiv preprint arXiv:2503.07588, 2025

    Pith/arXiv arXiv 2025

  9. [9]

    thinking with images

    Ziwei Zheng, Michael Yang, Jack Hong, Chenxiao Zhao, Guohai Xu, Le Yang, Chao Shen, and Xing Yu. Deepeyes: Incentivizing “thinking with images” via reinforcement learning.arXiv preprint arXiv:2505.14362, 2025

    Pith/arXiv arXiv 2025

  10. [10]

    Deepeyesv2: Toward agentic multimodal model.arXiv preprint arXiv:2511.05271, 2025

    Jack Hong, Chenxiao Zhao, ChengLin Zhu, Weiheng Lu, Guohai Xu, and Xing Yu. Deepeyesv2: Toward agentic multimodal model.arXiv preprint arXiv:2511.05271, 2025

    Pith/arXiv arXiv 2025

  11. [11]

    H. Shen, K. Zhao, T. Zhao, R. Xu, Z. Zhang, M. Zhu, and J. Yin. Zoomeye: Enhancing multimodal llms with human-like zooming capabilities through tree-based image exploration. InProceedings of the 2025 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 6613–6629, 2025

    2025

  12. [12]

    C. Wang, Z. Luo, R. Liu, C. Ran, S. Fan, X. Chen, and C. He. Vicot-agent: A vision-interleaved chain-of-thought framework for interpretable multimodal reasoning and scalable remote sensing analysis.arXiv preprint arXiv:2511.20085, 2025

    arXiv 2025

  13. [13]

    Zoomearth: Active perception for ultra-high-resolution geospatial vision-language tasks.arXiv preprint arXiv:2511.12267, 2025

    Ruixun Liu, Bowen Fu, Jiayi Song, Kaiyu Li, Wanchen Li, Lanxuan Xue, Hui Qiao, Weizhan Zhang, Deyu Meng, and Xiangyong Cao. Zoomearth: Active perception for ultra-high-resolution geospatial vision-language tasks.arXiv preprint arXiv:2511.12267, 2025

    arXiv 2025

  14. [14]

    Geoeyes: On-demand visual focusing for evidence-grounded understanding of ultra-high-resolution remote sensing imagery.arXiv preprint arXiv:2602.14201, 2026

    Fengxiang Wang, Mingshuo Chen, Yueying Li, Yajie Yang, Yifan Zhang, Long Lan, Xue Yang, Hongda Sun, Yulin Wang, Di Wang, et al. Geoeyes: On-demand visual focusing for evidence-grounded understanding of ultra-high-resolution remote sensing imagery.arXiv preprint arXiv:2602.14201, 2026

    arXiv 2026

  15. [15]

    Codev: Code with images for faithful visual reasoning via tool-aware policy optimization.arXiv preprint arXiv:2511.19661, 2025

    Xinhai Hou, Shaoyuan Xu, Manan Biyani, Mayan Li, Jia Liu, Todd C Hollon, and Bryan Wang. Codev: Code with images for faithful visual reasoning via tool-aware policy optimization.arXiv preprint arXiv:2511.19661, 2025

    arXiv 2025

  16. [16]

    Zooming without zooming: Region-to-image distillation for fine-grained multimodal perception.arXiv preprint arXiv:2602.11858, 2026

    Lai Wei, Liangbo He, Jun Lan, Lingzhong Dong, Yutong Cai, Siyuan Li, Huijia Zhu, Weiqiang Wang, Linghe Kong, Yue Wang, Zhuosheng Zhang, and Weiran Huang. Zooming without zooming: Region-to-image distillation for fine-grained multimodal perception.arXiv preprint arXiv:2602.11858, 2026

    arXiv 2026

  17. [17]

    What does vision tool-use reinforcement learning really learn? disentangling tool-induced and intrinsic effects for crop-and-zoom.arXiv preprint arXiv:2602.01334, 2026

    Yan Ma, Weiyu Zhang, Tianle Li, Linge Du, Xuyang Shen, and Pengfei Liu. What does vision tool-use reinforcement learning really learn? disentangling tool-induced and intrinsic effects for crop-and-zoom.arXiv preprint arXiv:2602.01334, 2026

    Pith/arXiv arXiv 2026

  18. [18]

    Reinforced attention learning.arXiv preprint arXiv:2602.04884, 2026

    Bangzheng Li, Jianmo Ni, Chen Qu, Ian Miao, Liu Yang, Xingyu Fu, Muhao Chen, and Derek Zhiyuan Cheng. Reinforced attention learning.arXiv preprint arXiv:2602.04884, 2026

    arXiv 2026

  19. [19]

    Smith, and Ranjay Krishna

    Yushi Hu, Weijia Shi, Xingyu Fu, Dan Roth, Mari Ostendorf, Luke Zettlemoyer, Noah A. Smith, and Ranjay Krishna. Visual sketchpad: Sketching as a visual chain of thought for multimodal language models. InAdvances in Neural Information Processing Systems, volume 37, pages 139348–139379, 2024. 12

    2024

  20. [20]

    Grounding dino: Marrying dino with grounded pre-training for open-set object detection.arXiv preprint arXiv:2303.05499, 2023

    Shilong Liu, Zhaoyang Zeng, Tianhe Ren, Feng Li, Hao Zhang, Jie Yang, Chunyuan Li, Jianwei Yang, Hang Su, Jun Zhu, et al. Grounding dino: Marrying dino with grounded pre-training for open-set object detection.arXiv preprint arXiv:2303.05499, 2023

    Pith/arXiv arXiv 2023

  21. [21]

    W. Ma, S. Sun, T. Yu, R. Wang, Tat-Seng Chua, and J. Bian. Thinking with blueprints: Assisting vision-language models in spatial reasoning via structured object representation, 2026

    2026

  22. [22]

    J. D. Silva, J. Magalhaes, D. Tuia, and B. Martins. Large language models for captioning and retrieving remote sensing images, 2024

    2024

  23. [23]

    Lhrs-bot: Em- powering remote sensing with vgi-enhanced large multimodal language model

    Dilxat Muhtar, Zhenshi Li, Feng Gu, Xueliang Zhang, and Pengfeng Xiao. Lhrs-bot: Em- powering remote sensing with vgi-enhanced large multimodal language model. InEuropean Conference on Computer Vision, pages 440–457. Springer, 2024

    2024

  24. [24]

    Lhrs-bot-nova: Improved multimodal large language model for remote sensing vision-language interpretation.arXiv preprint arXiv:2411.09301, 2024

    Zhenshi Li, Dilxat Muhtar, Feng Gu, Xueliang Zhang, Pengfeng Xiao, Guangjun He, and Xiaoxiang Zhu. Lhrs-bot-nova: Improved multimodal large language model for remote sensing vision-language interpretation.arXiv preprint arXiv:2411.09301, 2024

    Pith/arXiv arXiv 2024

  25. [25]

    Earthmind: Leveraging cross-sensor data for advanced earth observation interpretation with a unified multimodal llm.arXiv preprint arXiv:2506.01667, 2025

    Yan Shu, Bin Ren, Zhitong Xiong, Danda Pani Paudel, Luc Van Gool, Begum Demir, Nicu Sebe, and Paolo Rota. Earthmind: Leveraging cross-sensor data for advanced earth observation interpretation with a unified multimodal llm.arXiv preprint arXiv:2506.01667, 2025

    arXiv 2025

  26. [26]

    Klein, Salman Khan, and Fahad Khan

    Sagar Soni, Akshay Dudhane, Hiyam Debary, Mustansar Fiaz, Muhammad Akhtar Munir, Muhammad Sohail Danisho, Paolo Fraccaro, Campbell Watson, Levente J. Klein, Salman Khan, and Fahad Khan. Earthdial: Turning multi-sensory earth observations to interactive dialogues. ArXiv, 2025

    2025

  27. [27]

    J. Wang, Y . Zhong, Z. Chen, Z. Zheng, A. Ma, and L. Zhang. Earthvl: A progressive earth vision-language understanding and generation framework, 2026

    2026

  28. [28]

    P. Wang, H. Hu, B. Tong, Z. Zhang, F. Yao, Y . Feng, Z. Zhu, H. Chang, W. Diao, Q. Ye, and X. Sun. Ringmogpt: A unified remote sensing foundation model for vision, language, and grounded tasks.IEEE Transactions on Geoscience and Remote Sensing, 63:1–20, 2025

    2025

  29. [29]

    Earthmarker: A visual prompting multi-modal large language model for remote sensing.IEEE Transactions on Geoscience and Remote Sensing, 2024

    Wei Zhang, Miaoxin Cai, Tong Zhang, Yin Zhuang, Jun Li, and Xuerui Mao. Earthmarker: A visual prompting multi-modal large language model for remote sensing.IEEE Transactions on Geoscience and Remote Sensing, 2024

    2024

  30. [30]

    Rsunivlm: A unified vision-language model for remote sensing via granularity-oriented moe.Pattern Recognition, 179:113717, 2026

    Xu Liu and Zhouhui Lian. Rsunivlm: A unified vision-language model for remote sensing via granularity-oriented moe.Pattern Recognition, 179:113717, 2026

    2026

  31. [31]

    Bermano, and Ohad Fried

    Rotem Shalev-Arkushin, Rinon Gal, Amit H. Bermano, and Ohad Fried. Imagerag: Dynamic image retrieval for reference-guided image generation, 2025

    2025

  32. [32]

    Chatearthnet: A global- scale image-text dataset empowering vision-language geo-foundation models.Earth System Science Data Discussions, pages 1–24, 2024

    Zhenghang Yuan, Zhitong Xiong, Lichao Mou, and Xiao Xiang Zhu. Chatearthnet: A global- scale image-text dataset empowering vision-language geo-foundation models.Earth System Science Data Discussions, pages 1–24, 2024

    2024

  33. [33]

    Rsvqa: Visual question answering for remote sensing data.IEEE Transactions on Geoscience and Remote Sensing, 58(12):8555– 8566, 2020

    Sylvain Lobry, Diego Marcos, James Murray, and Devis Tuia. Rsvqa: Visual question answering for remote sensing data.IEEE Transactions on Geoscience and Remote Sensing, 58(12):8555– 8566, 2020

    2020

  34. [34]

    Mutual attention inception network for remote sensing visual question answering.IEEE Transactions on Geoscience and Remote Sensing, 60:1–14, 2021

    Xiangtao Zheng, Binqiang Wang, Xingqian Du, and Xiaoqiang Lu. Mutual attention inception network for remote sensing visual question answering.IEEE Transactions on Geoscience and Remote Sensing, 60:1–14, 2021

    2021

  35. [35]

    J. Wang, Z. Zheng, Z. Chen, A. Ma, and Y . Zhong. Earthvqa: Towards queryable earth via relational reasoning-based remote sensing visual question answering. InProceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 5481–5489, 2024

    2024

  36. [36]

    H2rsvlm: Towards helpful and honest remote sensing large vision language model.arXiv preprint arXiv:2403.20213, 2024

    Chao Pang, Jiang Wu, Jiayu Li, Yi Liu, Jiaxing Sun, Weijia Li, Xingxing Weng, Shuai Wang, Litong Feng, Gui-Song Xia, et al. H2rsvlm: Towards helpful and honest remote sensing large vision language model.arXiv preprint arXiv:2403.20213, 2024. 13

    Pith/arXiv arXiv 2024

  37. [37]

    Skysensegpt: A fine-grained instruction tuning dataset and model for remote sensing vision-language understanding.arXiv preprint arXiv:2406.10100, 2024

    Junwei Luo, Zhen Pang, Yongjun Zhang, Tingzhu Wang, Linlin Wang, Bo Dang, Jiangwei Lao, Jian Wang, Jingdong Chen, Yihua Tan, et al. Skysensegpt: A fine-grained instruction tuning dataset and model for remote sensing vision-language understanding.arXiv preprint arXiv:2406.10100, 2024

    Pith/arXiv arXiv 2024

  38. [38]

    Vhm: Versatile and honest vision language model for remote sensing image analysis

    Chao Pang, Xingxing Weng, Jiang Wu, Jiayu Li, Yi Liu, Jiaxing Sun, Weijia Li, Shuai Wang, Litong Feng, Gui-Song Xia, et al. Vhm: Versatile and honest vision language model for remote sensing image analysis. InProceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 6381–6388, 2025

    2025

  39. [39]

    Vrsbench: A versatile vision-language bench- mark dataset for remote sensing image understanding.arXiv preprint arXiv:2406.12384, 2024

    Xiang Li, Jian Ding, and Mohamed Elhoseiny. Vrsbench: A versatile vision-language bench- mark dataset for remote sensing image understanding.arXiv preprint arXiv:2406.12384, 2024

    Pith/arXiv arXiv 2024

  40. [40]

    Z. Su, P. Xia, H. Guo, Z. Liu, Y . Ma, X. Qu, J. Liu, Y . Li, K. Zeng, Z. Yang, et al. Thinking with images for multimodal reasoning: Foundations, methods, and future frontiers.arXiv preprint arXiv:2506.23918, 2025

    Pith/arXiv arXiv 2025

  41. [41]

    Visualtoolagent (vista): A reinforcement learning framework for visual tool selection.arXiv preprint arXiv:2505.20289, 2025

    Zeyi Huang, Yuyang Ji, Anirudh Sundara Rajan, Zefan Cai, Wen Xiao, Haohan Wang, Junjie Hu, and Yong Jae Lee. Visualtoolagent (vista): A reinforcement learning framework for visual tool selection.arXiv preprint arXiv:2505.20289, 2025

    Pith/arXiv arXiv 2025

  42. [42]

    Spacetools: Tool-augmented spatial reasoning via double interactive rl, 2025

    Siyi Chen, Mikaela Angelina Uy, Chan Hee Song, Faisal Ladhak, Adithyavairavan Murali, Qing Qu, Stan Birchfield, Valts Blukis, and Jonathan Tremblay. Spacetools: Tool-augmented spatial reasoning via double interactive rl, 2025

    2025

  43. [43]

    Reinforcing vlms to use tools for detailed visual reasoning under resource constraints.arXiv preprint arXiv:2506.14821, 2025

    Sunil Kumar, Bowen Zhao, Leo Dirac, and Paulina Varshavskaya. Reinforcing vlms to use tools for detailed visual reasoning under resource constraints.arXiv preprint arXiv:2506.14821, 2025

    Pith/arXiv arXiv 2025

  44. [44]

    Cropvlm: Learning to zoom for fine-grained vision-language perception.arXiv preprint arXiv:2511.19820, 2025

    Miguel Carvalho, Helder Dias, and Bruno Martins. Cropvlm: Learning to zoom for fine-grained vision-language perception.arXiv preprint arXiv:2511.19820, 2025

    Pith/arXiv arXiv 2025

  45. [45]

    Adaptive chain-of-focus reasoning via dynamic visual search and zooming for efficient vlms.arXiv preprint arXiv:2505.15436, 2025

    Xintong Zhang, Zhi Gao, Bofei Zhang, Pengxiang Li, Xiaowen Zhang, Yang Liu, Tao Yuan, Yuwei Wu, Yunde Jia, Song-Chun Zhu, and Qing Li. Adaptive chain-of-focus reasoning via dynamic visual search and zooming for efficient vlms.arXiv preprint arXiv:2505.15436, 2025

    Pith/arXiv arXiv 2025

  46. [46]

    Sensenova- mars: Empowering multimodal agentic reasoning and search via reinforcement learning.arXiv preprint arXiv:2512.24330, 2025

    Yong Xien Chng, Tao Hu, Wenwen Tong, Xueheng Li, Jiandong Chen, Haojia Yu, Jiefan Lu, Hewei Guo, Hanming Deng, Chengjun Xie, Gao Huang, Dahua Lin, and Lewei Lu. Sensenova- mars: Empowering multimodal agentic reasoning and search via reinforcement learning.arXiv preprint arXiv:2512.24330, 2025

    arXiv 2025

  47. [47]

    Pixel reasoner: Incentivizing pixel-space reasoning with curiosity-driven reinforcement learning.arXiv preprint arXiv:2505.15966, 2025

    Alex Su, Haozhe Wang, Weiming Ren, Fangzhen Lin, and Wenhu Chen. Pixel reasoner: Incentivizing pixel-space reasoning with curiosity-driven reinforcement learning.arXiv preprint arXiv:2505.15966, 2025

    Pith/arXiv arXiv 2025

  48. [48]

    Reinforcing spatial reasoning in vision-language models with interwoven thinking and visual drawing, 2025

    Junfei Wu, Jian Guan, Kaituo Feng, Qiang Liu, Shu Wu, Liang Wang, Wei Wu, and Tieniu Tan. Reinforcing spatial reasoning in vision-language models with interwoven thinking and visual drawing, 2025

    2025

  49. [49]

    Shuai Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, Sibo Song, Kai Dang, Peng Wang, Shijie Wang, Jun Tang, et al. Qwen2. 5-vl technical report.arXiv preprint arXiv:2502.13923, 2025

    Pith/arXiv arXiv 2025

  50. [50]

    Introducing gpt-5.4

    OpenAI. Introducing gpt-5.4. https://openai.com/zh-Hans-CN/index/ introducing-gpt-5-4/, 2026. Accessed: 2026-05-06

    2026

  51. [51]

    The claude 4.6 model family

    Anthropic. The claude 4.6 model family. Technical report, Anthropic, 2026

    2026

  52. [52]

    Qwen3-vl technical report.arXiv preprint arXiv:2511.21631, 2025

    Shuai Bai, Yuxuan Cai, Ruizhe Chen, Keqin Chen, Xionghui Chen, Zesen Cheng, Lianghao Deng, Wei Ding, Chang Gao, Chunjiang Ge, Wenbin Ge, Zhifang Guo, Qidong Huang, Jie Huang, Fei Huang, Binyuan Hui, Shutong Jiang, Zhaohai Li, Mingsheng Li, Mei Li, Kaixin Li, Zicheng Lin, Junyang Lin, Xuejing Liu, Jiawei Liu, Chenglong Liu, Yang Liu, Dayiheng Liu, Shixuan ...

    Pith/arXiv arXiv 2025

  53. [53]

    Gpt-5.2: Advancing science and math

    OpenAI. Gpt-5.2: Advancing science and math. https://openai.com/zh-Hans-CN/ index/introducing-gpt-5-2/, 2025. Accessed: 2026-05-06

    2025

  54. [54]

    Cogcom: Train large vision-language models diving into details through chain of manipulations.arXiv preprint arXiv:2402.04236, 2024

    Ji Qi, Ming Ding, Weihan Wang, Yushi Bai, Qingsong Lv, Wenyi Hong, Bin Xu, Lei Hou, Juanzi Li, Yuxiao Dong, and Jie Tang. Cogcom: Train large vision-language models diving into details through chain of manipulations.arXiv preprint arXiv:2402.04236, 2024

    Pith/arXiv arXiv 2024

  55. [55]

    Deepseekmath: Pushing the limits of mathematical reasoning in open language models.arXiv preprint arXiv:2402.03300, 2024

    DeepSeek-AI. Deepseekmath: Pushing the limits of mathematical reasoning in open language models.arXiv preprint arXiv:2402.03300, 2024

    Pith/arXiv arXiv 2024

  56. [56]

    Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning.arXiv preprint arXiv:2501.12948, 2025

    DeepSeek-AI. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning.arXiv preprint arXiv:2501.12948, 2025

    Pith/arXiv arXiv 2025

  57. [57]

    Claude 3.7 sonnet.https://www.anthropic.com, 2025

    Anthropic. Claude 3.7 sonnet.https://www.anthropic.com, 2025. Accessed: 2026-05

    2025

  58. [58]

    Gemini 2.5: Pushing the frontier with advanced reasoning, multimodality, long context, and agentic capabilities.arXiv preprint arXiv:2507.06261, 2025

    Google DeepMind. Gemini 2.5: Pushing the frontier with advanced reasoning, multimodality, long context, and agentic capabilities.arXiv preprint arXiv:2507.06261, 2025

    Pith/arXiv arXiv 2025

  59. [59]

    Internvl3: Exploring advanced training and test-time recipes for open-source multimodal models.arXiv preprint arXiv:2504.10479, 2025

    Jinguo Zhu, Weiyun Wang, Zhe Chen, Zhaoyang Liu, Shenglong Ye, Lixin Gu, Yuchen Duan, Hao Tian, Weijie Su, Jie Shao, et al. Internvl3: Exploring advanced training and test-time recipes for open-source multimodal models.arXiv preprint arXiv:2504.10479, 2025

    Pith/arXiv arXiv 2025

  60. [60]

    Qwen2-vl: Enhancing vision-language model’s perception of the world at any resolution.arXiv preprint arXiv:2409.12191, 2024

    Peng Wang, Shuai Bai, Sinan Tan, Shijie Wang, Zhihao Fan, Jinze Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, et al. Qwen2-vl: Enhancing vision-language model’s perception of the world at any resolution.arXiv preprint arXiv:2409.12191, 2024

    Pith/arXiv arXiv 2024

  61. [61]

    Internvl 2.5: Scaling up vision-language models with enhanced visual encoding

    OpenGVLab. Internvl 2.5: Scaling up vision-language models with enhanced visual encoding. https://internvl.github.io, 2024. Technical report

    2024

  62. [62]

    Intern-s1-mini

    OpenGVLab. Intern-s1-mini. https://github.com/OpenGVLab, 2024. Lightweight multi- modal model

    2024

  63. [63]

    Scaling vision pre-training to 4k resolution, 2025

    Baifeng Shi, Boyi Li, Han Cai, Yao Lu, Sifei Liu, Marco Pavone, Jan Kautz, Song Han, Trevor Darrell, Pavlo Molchanov, and Hongxu Yin. Scaling vision pre-training to 4k resolution, 2025

    2025

  64. [64]

    Minicpm-v: A gpt-4v level mllm on your phone.arXiv preprint arXiv:2408.01800, 2024

    Yuan Yao, Tianyu Yu, Ao Zhang, Chongyi Wang, Junbo Cui, Hongji Zhu, Tianchi Cai, Haoyu Li, Weilin Zhao, Zhihui He, et al. Minicpm-v: A gpt-4v level mllm on your phone.arXiv preprint arXiv:2408.01800, 2024

    Pith/arXiv arXiv 2024

  65. [65]

    Internvl3.5: Advancing open-source multimodal models in versatility, reasoning, and efficiency.arXiv preprint arXiv:2508.18265, 2025

    Weiyun Wang, Zhangwei Gao, Lixin Gu, Hengjun Pu, Long Cui, Xingguang Wei, Zhaoyang Liu, Linglin Jing, Shenglong Ye, Jie Shao, et al. Internvl3.5: Advancing open-source multimodal models in versatility, reasoning, and efficiency.arXiv preprint arXiv:2508.18265, 2025

    Pith/arXiv arXiv 2025

  66. [66]

    Phi-3 technical report: A highly capable language model locally on your phone.arXiv preprint arXiv:2404.14219, 2024

    Microsoft GenAI Team. Phi-3 technical report: A highly capable language model locally on your phone.arXiv preprint arXiv:2404.14219, 2024

    Pith/arXiv arXiv 2024

  67. [67]

    Deepseek-vl: towards real-world vision-language understanding.arXiv preprint arXiv:2403.05525, 2024

    Haoyu Lu, Wen Liu, Bo Zhang, Bingxuan Wang, Kai Dong, Bo Liu, Jingxiang Sun, Tongzheng Ren, Zhuoshu Li, Yaofeng Sun, et al. Deepseek-vl: towards real-world vision-language understanding.arXiv preprint arXiv:2403.05525, 2024

    Pith/arXiv arXiv 2024

  68. [68]

    Hello gpt-4o.https://openai.com/index/hello-gpt-4o, 2024

    OpenAI. Hello gpt-4o.https://openai.com/index/hello-gpt-4o, 2024

    2024

  69. [69]

    Gpt-4o mini: advancing cost-efficient intelligence

    OpenAI. Gpt-4o mini: advancing cost-efficient intelligence. https://openai.com/index/ gpt-4o-mini-advancing-cost-efficient-intelligence, 2024

    2024

  70. [70]

    Llava-onevision-1.5: Fully open framework for democratized multimodal training.arXiv preprint arXiv:2509.23661, 2025

    Xiang An, Yin Xie, Kaicheng Yang, et al. Llava-onevision-1.5: Fully open framework for democratized multimodal training.arXiv preprint arXiv:2509.23661, 2025. 15

    Pith/arXiv arXiv 2025

  71. [71]

    Qwen3 technical report.arXiv preprint arXiv:2505.09388, 2025

    An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, Chujie Zheng, Dayiheng Liu, Fan Zhou, Fei Huang, Feng Hu, Hao Ge, Haoran Wei, Huan Lin, Jialong Tang, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxi Yang, Jing Zhou, Jingren Zhou, Junyang Lin, Kai Dang, Keqin Bao, Kexin Yang, ...

    Pith/arXiv arXiv 2025

  72. [72]

    Internlm-xcomposer-2.5: A versatile large vision language model supporting long-contextual input and output.arXiv preprint arXiv:2407.03320, 2024

    Pan Zhang, Xiaoyi Dong, Yuhang Zang, Yuhang Cao, Rui Qian, Lin Chen, Qipeng Guo, Haodong Duan, Bin Wang, Linke Ouyang, et al. Internlm-xcomposer-2.5: A versatile large vision language model supporting long-contextual input and output.arXiv preprint arXiv:2407.03320, 2024

    Pith/arXiv arXiv 2024

  73. [73]

    Internlm2 technical report

    Shanghai AI Laboratory. Internlm2 technical report. https://github.com/InternLM, 2023. Open-source LLM

    2023

  74. [74]

    Internlm3.https://github.com/InternLM, 2024

    Shanghai AI Laboratory. Internlm3.https://github.com/InternLM, 2024. Next-generation InternLM series

    2024

  75. [75]

    Vicuna: An open-source chatbot impressing gpt-4 with 90% chatgpt quality.arXiv preprint arXiv:2306.05685, 2023

    Wei-Lin Chiang, Lianmin Zheng, Ying Sheng, et al. Vicuna: An open-source chatbot impressing gpt-4 with 90% chatgpt quality.arXiv preprint arXiv:2306.05685, 2023

    Pith/arXiv arXiv 2023

  76. [76]

    Vlm-r 3: Region recognition, reasoning, and refinement for enhanced multimodal chain-of-thought.arXiv preprint arXiv:2505.16192, 2025

    Chaoya Jiang, Yongrui Heng, Wei Ye, Han Yang, Haiyang Xu, Ming Yan, Ji Zhang, Fei Huang, and Shikun Zhang. Vlm-r 3: Region recognition, reasoning, and refinement for enhanced multimodal chain-of-thought.arXiv preprint arXiv:2505.16192, 2025

    Pith/arXiv arXiv 2025

  77. [77]

    Fair1m: A benchmark dataset for fine-grained object recognition in high-resolution remote sensing imagery.ISPRS Journal of Photogrammetry and Remote Sensing, 184:116–130, 2022

    Xian Sun, Peijin Wang, Zhiyuan Yan, Feng Xu, Ruiping Wang, Wenhui Diao, Jin Chen, Jihao Li, Yingchao Feng, Tao Xu, et al. Fair1m: A benchmark dataset for fine-grained object recognition in high-resolution remote sensing imagery.ISPRS Journal of Photogrammetry and Remote Sensing, 184:116–130, 2022

    2022

  78. [78]

    Yansheng Li, Linlin Wang, Tingzhu Wang, Xue Yang, Junwei Luo, Qi Wang, Youming Deng, Wenbin Wang, Xian Sun, Haifeng Li, et al. Star: A first-ever dataset and a large-scale benchmark for scene graph generation in large-size satellite imagery.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2(5):6, 2024. 16 A Appendix A.1 Overview of the Appe...

    2024