Pith. sign in

REVIEW 2 major objections 5 minor 49 references

LookAgain: Closed-Loop GUI Grounding with Visually Grounded Reflection

T0 review · 2 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read LookAgain reformulates GUI grounding as a closed predict-look-again-refine loop, treating each emitted coordinate as a hypothesis to be verified and revised, and reports large accuracy gains over one-shot grounders.

desk verdict A genuinely new closed-loop GUI-grounding protocol with strong results, but the paper's key ablation confounds the multi-turn mechanism with SFT on Gemini-distilled trajectories. read the letter →

arxiv 2608.09723 v1 pith:WACIELRD submitted 2026-08-10 cs.CV

Renshan Zhang , Haoyang Meng , Yixiao He , Rui Shao , April Hua Liu , Liqiang Nie This is my paper · ORCID
classification cs.CV
keywords GUIgroundingclosed-looprefinementvisuallygroundedreflectionmulti-turnrefusal-awareGRPOscreenagents
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 claims that the main bottleneck in GUI grounding is not what happens before a click but what happens after it: existing grounders freeze a predicted coordinate at the moment it is emitted, so no mechanism can challenge it under new visual evidence. LookAgain reformulates grounding as a closed predict-look-again-refine loop, where a 'locate' call posts a coordinate, renders a red marker at that point, and appends a high-resolution patch of the region, and a 'confirm' call either commits the coordinate or refuses the instruction. On top of a base 8B vision-language model, the closed loop raises OSWorld-G overall accuracy from 51.3 to 73.0, with the largest gains on small targets, dense layouts, and refusal cases where the instruction does not apply. The paper's ablation isolates the source of the gain: a single-turn reinforcement-learning variant reaches only 47.6 on ScreenSpot-Pro, while the full multi-turn reflection model reaches 60.2. The claim matters because grounding errors are a dominant cause of cascading failures in screen-operating agents, and this result suggests grounding should be treated as an evidence-accumulating decision rather than a one-shot regression.

What carries the argument

The load-bearing mechanism is the closed predict-look-again-refine loop built from two tool-use primitives. 'locate' renders a red marker at the predicted center and appends a 512×512 patch of that exact region, making the model's own hypothesis the addressable subject of the next reasoning step; 'confirm' commits or refuses and terminates. Around these primitives, the training recipe is the second half of the machinery: a turn-selective SFT mask that supervises genuine post-hoc reflections but never imitates a wrong intermediate coordinate, and a GRPO stage whose only reward is terminal grounding correctness, gated on a successful confirm and zeroed out on malformed calls, repeated locate boxes, or truncated trajectories.

What would settle it

Evaluate LookAgain on refusal cases where the absent element is not covered by a same-image patch — for example, instructions referring to a button that never existed in a crowded toolbar. If refusal accuracy on this set is close to the synthetic refusal accuracy, the model understands absence; if it collapses while accuracy on the feathered-adjacent-patch cases stays high, the model has learned compositing artifacts rather than absence. A cheaper check is to remove the feathering from the synthetic refusal images: if accuracy drops sharply, the seam is the cue.

Watch

Extended reading notes

Core claim

LookAgain's central claim is that a produced coordinate should be treated as a hypothesis to be reflected upon and revised under new visual evidence. The method implements this as a multi-turn tool-use protocol with two primitives: 'locate' posts a bounding-box hypothesis, overlays a red-dot-and-ring marker at its center, and appends a 512×512 crop of that exact region to the dialogue history; 'confirm' accepts or rejects the hypothesis and ends the episode, with 'failed' status used to abstain on inapplicable instructions. Every subsequent thought is therefore conditioned on the model's own previous prediction and the visual patch it summoned, turning iterative zoom-in over views into iterative correction over coordinates. Training is a two-stage pipeline: supervised fine-tuning on teacher-generated reflective trajectories (with a selective token mask that avoids imitating wrong intermediate coordinates), followed by GRPO whose only reward is whether the final confirm-committed prediction hits the ground-truth box (or, on refusal samples, whether the model correctly declares the instruction inapplicable). The paper reports that the 8B model reaches 73.0 overall on OSWorld-G versus 51.3 for the base model, 84.5 on MMBench-L2-GUI, 60.2 on ScreenSpot-Pro, and 38.8 on UIVision, and that the multi-turn reflection mechanism — not reinforcement learning alone — is what produces the gain.

Load-bearing premise

The refusal behavior is trained almost entirely on synthetic 'the element is missing' images made by pasting a feathered patch from an adjacent side over the target, and the reported refusal accuracy transfers to real inapplicable instructions only if those composite images are a faithful stand-in for genuine absences.

Editorial extensions

If this is right

  • Grounding errors on small targets and dense layouts, a dominant source of cascading failures in screen agents, can be reduced by letting the model inspect its own predicted coordinate.
  • Refusal awareness — correctly deciding an instruction is inapplicable — can be trained, and on OSWorld-G the refusal subset jumps from 0.0 for several strong baselines to 57.4 for the 8B model without sacrificing standard accuracy.
  • The multi-turn reflection gain is not an artifact of reinforcement learning alone: a single-turn GRPO variant reaches only 47.6 on ScreenSpot-Pro versus 60.2 for LookAgain.
  • The same recipe transfers across backbones; on a 9B model it raises OSWorld-G-Refine to 84.9, UIVision to 49.9, and VenusBench-GD to 72.1.

Reading between the lines

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

  • A natural testable extension is to run the same locate-confirm loop on other fine-grained localization domains — document extraction, medical imaging, remote sensing — where a coordinate hypothesis can be rendered and inspected; the paper does not claim this, but the mechanism is domain-agnostic.
  • The refusal numbers depend on a CutPaste-style synthetic distribution where the absent element is always erased by pasting an adjacent patch over it; real inapplicable instructions often lack such a same-image patch to paste, so transfer of the reported 57.4 refusal accuracy to organic refusal cases remains an open question.
  • If verification is separated from commitment, evaluation should start reporting budgeted multi-turn grounding accuracy — for example, accuracy at a fixed number of locate calls — rather than single-shot accuracy alone, since the paper's bbox-count analysis suggests such budgets would trade accuracy against latency.
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

2 major / 5 minor

Summary. The paper proposes LookAgain, a closed-loop GUI grounding method that reformulates grounding as a multi-turn predict–look-again–refine process. A 'locate' tool posts a coordinate hypothesis, renders a red marker on the screenshot, and appends a local patch around the predicted region; a 'confirm' tool accepts or rejects the hypothesis and terminates the episode. Training is two-stage: SFT on roughly 70k Gemini3.1-Pro-distilled reflective trajectories, followed by GRPO with a terminal grounding correctness reward, with additional refusal cases synthesized via a CutPaste-style heuristic. Experiments on OSWorld-G, VenusBench-GD, ScreenSpot-Pro, UIVision, and MMBench-L2-GUI report large gains over the base Qwen3-VL model and state-of-the-art results. Ablations in Tables 3–5 attempt to attribute the gains to the multi-turn reflection mechanism and show generalization to a second backbone, but the central attribution is complicated by the SFT/RL comparison.

Significance. The contribution is conceptually clean and potentially impactful: it introduces a simple, general protocol for post-prediction visual verification, and the empirical gains are substantial and consistent across several benchmarks, including a new refusal-aware capability. The paper ships code, uses a rule-based reward, and provides multiple ablations, backbone generalization, and an analysis of refinement dynamics. If the mechanistic attribution is confirmed, this would be a useful recipe for the community. However, the current comparison does not cleanly separate the multi-turn protocol from the teacher-distilled SFT cold-start, and the refusal behavior is trained on synthetic composites whose validity is unverified; both issues need to be resolved before the central claims can be accepted.

major comments (2)
  1. [Ablation Study, Table 3] The comparison in Table 3 between LookAgain-8B and the '+GRPO-Single Turn' variants does not isolate the multi-turn reflection mechanism because LookAgain-8B includes an SFT cold-start on roughly 70k Gemini3.1-Pro-distilled trajectories, while the single-turn GRPO variants are trained from the base model with reinforcement learning only. Table 4 shows that SFT alone raises UIVision from 23.3 to 37.5 (close to LookAgain's 38.8) and OSWorld-G-Refine from 67.8 to 73.5, so the residual gap in Table 3 (e.g., 60.2 versus 47.6/55.7 on ScreenSpot-Pro) could be substantially explained by teacher distillation rather than by the closed-loop protocol. The statement 'showing that the gain stems from multi-turn reflection itself rather than from simply avoiding the refusal burden' is therefore not supported by the reported experiments. Please add an ablation with a single-turn SFT+GRPO variant (or a multi-turn GRPO variant without SFT) to separate these factors.
  2. [Appendix, Refusal Data Construction] The refusal training data is synthesized by pasting a patch from an adjacent side over the ground-truth bounding box and feathering the boundary (CutPaste-style). With only about 600 refusal trajectories for SFT and 300 for RL, the model may learn to detect composite-specific artifacts (seams, texture discontinuities) rather than to reason that the instruction's referent is absent. This is a particular concern because the paper reports large refusal gains over baselines that score 0.0 (e.g., 57.4 on OSWorld-G Refusal), and the benchmarks' refusal cases are presumably natural inapplicable instructions. The paper provides no validation that the synthetic refusal distribution matches real refusal scenarios. Please include a human evaluation of the synthetic composites (e.g., how often they look like genuine screenshots) and/or an evaluation on a held-out set of naturally occurring inapplicable instructions.
minor comments (5)
  1. [Introduction] There is a typo in the first paragraph: 'degradation sharply small targets' is missing 'on' (should be 'degradation sharply on small targets').
  2. [Implementation Details, Eq. (3)] The locate repetition limit K used in the Violation condition is not specified anywhere; please report its value alongside the maximum interaction rounds.
  3. [Method, Figure 3] Figure 3 is referenced in the Analysis of Direct Closed-Loop Grounding section, but the quantitative scores for the direct closed-loop application to Gemini3.1-Pro, Qwen3-VL-8B, and the GUI-Cursor variant are not given in the text; please report the exact numbers.
  4. [Tables 1 and 3] The benchmark name is inconsistent: 'VenusBench-GD' in Table 1 caption and 'Venus-GD' in Table 3; please standardize.
  5. [Implementation Details] The statement 'All evaluations are repeated three times to ensure negligible fluctuation in results' should be accompanied by the actual variance (standard deviation or range) in the main tables; otherwise the 'negligible' claim is unverifiable.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the training targets, rewards, and evaluation benchmarks are all external to the claimed predictions.

full rationale

The paper's derivation chain is: define a closed-loop locate/confirm protocol, distill ~70k reflective trajectories from Gemini3.1-Pro filtered by ground-truth bounding boxes, cold-start with SFT, refine with GRPO under an outcome reward (Eqs. 2-4) gated on ground-truth containment or refusal labels, and evaluate on external benchmarks (OSWorld-G, VenusBench-GD, ScreenSpot-Pro, UIVision, MMBench-L2-GUI). None of these stages defines the evaluation target in terms of the model's own outputs. The SFT corpus is filtered by GT correctness, the RL reward uses GT boxes, and the reported numbers are measured on held-out benchmark subsets; there is no fitted parameter that is renamed as a prediction and no equation in which the predicted quantity equals the training input by construction. The refusal synthesis (CutPaste-style overwriting) is a training-data construction and does not supply the labels for the external refusal benchmarks, so the refusal results are not self-definitional. The self-citations in Related Work are contextual and not load-bearing; no uniqueness theorem or prior result is invoked to force the design. The main weakness is experimental rather than circular: Table 3 compares the full SFT+GRPO multi-turn model with single-turn GRPO variants that lack the SFT cold-start, so the incremental effect attributed to multi-turn reflection is underdetermined by that comparison, and Table 4 shows SFT alone accounts for much of the gain. This is an attribution and confound concern, not a reduction of the prediction to its inputs.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical entities or exotic constructs. Its contributions are a training protocol and tool-use format. The free parameters are training hyperparameters; the axioms are empirical assumptions about the training and evaluation setup.

free parameters (6)
  • reward coefficient alpha = 0.1
    Weight for format reward in GRPO (Eq. 4). Chosen by hand, not tuned across values.
  • reward coefficient beta = 0.9
    Weight for hit reward in GRPO (Eq. 4). Chosen by hand.
  • maximum interaction rounds = 6
    Turn limit for RL training and evaluation; affects how many refinements are possible.
  • locate repetition limit K = not specified
    Violation condition (ii) in Eq. (3): same locate bbox emitted more than K times is penalized. K is not given a numeric value in the text.
  • patch size = 512x512
    Resolution of the cropped local patch in the locate primitive.
  • coordinate scale = 0-1000
    Normalized coordinate range used for locate calls; standard for GUI grounding.
assumptions (4)
  • domain assumption Marker-annotated local patch provides sufficient visual evidence for the model to verify or correct its previous prediction
    The entire closed-loop mechanism (Method: The locate primitive) assumes that rendering a red dot/ring and cropping a 512x512 patch around the predicted point lets the model see whether it hit the target. This is validated empirically but not proven.
  • domain assumption Filtered teacher trajectories are a sound training signal for transferring reflection behavior
    Stage 1 SFT assumes that Gemini 3.1 Pro trajectories, filtered by final correctness, teach the student genuine reflect-and-refine behavior rather than superficial imitation. Figure 10(c) suggests GUI-Cursor style RL alone fails, but the SFT assumption is still a premise.
  • domain assumption CutPaste refusal synthesis creates valid refusal cases without exploitable artifacts
    Refusal data construction assumes pasting a neighboring patch over the target and feathering yields a sample where the instruction is genuinely inapplicable, with no detectable seam. No validation is provided.
  • domain assumption GRPO with terminal correctness reward can shape multi-turn behavior without reward hacking
    Stage 2 assumes the rule-based reward plus violation penalties is sufficient to prevent degenerate loops (malformed calls, repetition, truncation) and encourage genuine refinement.

how reviews work

0 comments
Cite this review

Pith. "Pith review of LookAgain: Closed-Loop GUI Grounding with Visually Grounded Reflection." pith.science (2026). https://pith.science/paper/WACIELRD

@misc{pith2026260809723,
  author       = {Pith},
  title        = {Pith review of: LookAgain: Closed-Loop GUI Grounding with Visually Grounded Reflection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WACIELRD}},
  note         = {Machine review of arXiv:2608.09723}
}
read the original abstract

Recent graphical user interface (GUI) grounders have significantly advanced single-shot accuracy on standard benchmarks, yet their performance degrades sharply on small targets, densely packed controls and out-of-distribution interfaces. We attribute this gap to a paradigmatic limitation shared by existing approaches: none of them treats a produced coordinate as a hypothesis to be reflected upon and revised under new visual evidence. This manifests as three coupled issues: 1) Lack of post-hoc reflection. The prediction is frozen at the moment of emission, leaving no internal mechanism to challenge or refine it. 2) Visual evidence decoupled from the prediction. The auxiliary visual evidence is gathered to support the upcoming coordinate rather than to scrutinise the one already committed to. 3) Refinement over views, not over predictions. The iterative zoom-in refines the inspected region instead of inheriting a previous coordinate as a spatial prior to be corrected. In this paper, we propose LookAgain, a closed-loop GUI grounder driven by post-prediction visual reflection. LookAgain reformulates grounding as a multi-turn predict-look-again-refine process with two primitives: "locate" posts a coordinate hypothesis, renders a marker on the image and appends a local patch of the predicted region. It anchors the next reasoning step to the previous prediction as a spatial prior; "confirm" accepts or reject the hypothesis and terminates the procedure. We train the LookAgain grounder with SFT on constructed reflective trajectories as a cold start, followed by GRPO with terminal grounding correctness as the sole reward. Extensive experiments show that LookAgain consistently improves performance on both refusal-aware and general GUI grounding benchmarks, achieving state-of-the-art results. Comprehensive ablations further verify the effectiveness of the proposed framework.

Figures

Figures reproduced from arXiv: 2608.09723 by the authors.

Figure 1
Figure 1. Comparison between existing GUI grounding [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overall training framework of LookAgain. The model is first cold-started by supervised fine-tuning on teacher [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Direct application of closed-loop grounding across [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Relationship between target size and predic￾tion count in LookAgain. The left figure shows the average number of predicted bounding boxes for each benchmark, to￾gether with the corresponding average target area ratio. The right figure shows the average number of predic…
Figure 5
Figure 5. Figure 5: Distribution of the number of predicted bounding boxes across different benchmarks. Each subplot shows the distribution of bounding box prediction counts on one benchmark, with the corresponding target area ratio reported in the title. Consistent with the main discussi…
Figure 6
Figure 6. Figure 6: Training dynamics of LookAgain. From left to right, the three plots present the training reward, the validation accuracy, and the average number of predicted bounding boxes on the validation set, respectively. The synchronized increase in validation accuracy and predic…
Figure 7
Figure 7. Figure 7: A qualitative example illustrating the reflection and refinement capability of LookAgain. The case shows that LookAgain progressively revises its previous predictions by identifying incorrect toolbar icons and refining the grounding result over multiple turns. This exa…
Figure 8
Figure 8. Figure 8: System prompt used by LookAgain. Method OSG OSG-R UIV Propose-then-Critic-8B (Wang et al. 2026a) 59.6 - 28.5 Qwen3-VL-8B w/ MVP (Zhang et al. 2026) - 72.7 31.9 LookAgain-8B 73.0 81.0 38.8 [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]
Figure 9
Figure 9. Figure 9: Tools definition used by the grounding agent. indicating that a critic-style selection step is less effective than iterative, visually grounded revision. Qualitative Results Case Study We provide a qualitative example in [PITH_FULL_IMAGE:figures/full_fig_p013_9.png]
Figure 10
Figure 10. Figure 10: Reflection trajectory comparison of four grounders. Turn-by-turn trajectories of (a) Qwen3-VL-8B-Instruct, (b) Gemini3.1-Pro w/ closed-loop protocol, (c) the GUI-Cursor variant, and (d) LookAgain-8B, showing that only Gemini3.1-Pro and LookAgain-8B perform genuine ref…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

49 extracted references · 30 canonical work pages

  1. [1]

    Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=

    Seeclick: Harnessing gui grounding for advanced visual gui agents , author=. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=

  2. [2]

    International Conference on Learning Representations , volume=

    OS-ATLAS: Foundation action model for generalist GUI agents , author=. International Conference on Learning Representations , volume=

    work page

  3. [3]

    2026 , url=

    Liangyu Chen and Hanzhang Zhou and chenglin cai and Jianan Zhang and Panrong Tong and Xu Zhang and Quyu Kong and Chen Liu and Yuqi Liu and Wenxuan Wang and Yue Wang and Qin Jin and Steven HOI , booktitle=. 2026 , url=

    work page 2026

  4. [4]

    arXiv preprint arXiv:2504.10458 , year=

    Gui-r1: A generalist r1-style vision-language action model for gui agents , author=. arXiv preprint arXiv:2504.10458 , year=

    arXiv

  5. [5]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Ui-r1: Enhancing efficient action prediction of gui agents by reinforcement learning , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  6. [6]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    GUI-G ^2 : Gaussian Reward Modeling for GUI Grounding , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    work page

  7. [7]

    Advances in Neural Information Processing Systems , volume=

    Gui-g1: Understanding r1-zero-like training for visual grounding in gui agents , author=. Advances in Neural Information Processing Systems , volume=

    work page

  8. [8]

    Advances in Neural Information Processing Systems , volume=

    Gui-actor: Coordinate-free visual grounding for gui agents , author=. Advances in Neural Information Processing Systems , volume=

    work page

Show all 49 references
  1. [9]

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

    Mvp: Multiple view prediction improves gui grounding , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  2. [10]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Visual test-time scaling for gui agent grounding , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  3. [11]

    Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing , pages=

    Dimo-gui: Advancing test-time scaling in gui grounding via modality-aware visual reasoning , author=. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing , pages=

    2025

  4. [12]

    Advances in neural information processing systems , volume=

    Reflexion: Language agents with verbal reinforcement learning , author=. Advances in neural information processing systems , volume=

  5. [13]

    International Conference on Learning Representations , volume=

    Navigating the digital world as humans do: Universal visual grounding for gui agents , author=. International Conference on Learning Representations , volume=

  6. [14]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Showui: One vision-language-action model for gui visual agent , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  7. [15]

    Findings of the Association for Computational Linguistics: ACL 2025 , pages=

    Aria-ui: Visual grounding for gui instructions , author=. Findings of the Association for Computational Linguistics: ACL 2025 , pages=

    2025

  8. [16]

    Proceedings of the 33rd ACM International Conference on Multimedia , pages=

    Screenspot-pro: Gui grounding for professional high-resolution computer use , author=. Proceedings of the 33rd ACM International Conference on Multimedia , pages=

  9. [17]

    International Conference on Machine Learning , pages=

    UI-Vision: A Desktop-centric GUI Benchmark for Visual Perception and Interaction , author=. International Conference on Machine Learning , pages=. 2025 , organization=

    2025

  10. [18]

    arXiv preprint arXiv:2504.14239 , year=

    Infigui-r1: Advancing multimodal gui agents from reactive actors to deliberative reasoners , author=. arXiv preprint arXiv:2504.14239 , year=

    arXiv

  11. [19]

    arXiv preprint arXiv:2501.12948 , year=

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

    arXiv

  12. [20]

    arXiv preprint arXiv:2601.06899 , year=

    V2P: Visual Attention Calibration for GUI Grounding via Background Suppression and Center Peaking , author=. arXiv preprint arXiv:2601.06899 , year=

  13. [21]

    arXiv preprint arXiv:2511.21631 , year=

    Qwen3-vl technical report , author=. arXiv preprint arXiv:2511.21631 , year=

    arXiv

  14. [22]

    arXiv preprint arXiv:2507.05791 , year=

    Gta1: Gui test-time scaling agent , author=. arXiv preprint arXiv:2507.05791 , year=

    arXiv

  15. [23]

    Advances in Neural Information Processing Systems , volume=

    Opencua: Open foundations for computer-use agents , author=. Advances in Neural Information Processing Systems , volume=

  16. [24]

    arXiv preprint arXiv:2508.10833 , year=

    Ui-venus technical report: Building high-performance ui agents with rft , author=. arXiv preprint arXiv:2508.10833 , year=

    arXiv

  17. [25]

    2025 , url=

    Holo1.5 - Open Foundation Models for Computer Use Agents , author=. 2025 , url=

    2025

  18. [26]

    arXiv preprint arXiv:2501.12326 , year=

    Ui-tars: Pioneering automated gui interaction with native agents , author=. arXiv preprint arXiv:2501.12326 , year=

    arXiv

  19. [27]

    Holo2 - Open Foundation Models for Navigation and Computer Use Agents , author=

  20. [28]

    arXiv preprint arXiv:2508.15144 , year=

    Mobile-agent-v3: Fundamental agents for gui automation , author=. arXiv preprint arXiv:2508.15144 , year=

    arXiv

  21. [29]

    Advances in Neural Information Processing Systems , volume=

    Scaling computer-use grounding via user interface decomposition and synthesis , author=. Advances in Neural Information Processing Systems , volume=

  22. [30]

    arXiv preprint arXiv:2512.16501 , year=

    VenusBench-GD: A Comprehensive Multi-Platform GUI Benchmark for Diverse Grounding Tasks , author=. arXiv preprint arXiv:2512.16501 , year=

  23. [31]

    arXiv preprint arXiv:2507.19478 , year=

    Mmbench-gui: Hierarchical multi-platform evaluation framework for gui agents , author=. arXiv preprint arXiv:2507.19478 , year=

    arXiv

  24. [32]

    arXiv preprint arXiv:2606.14579 , year=

    VISTA: View-Consistent Self-Verified Training for GUI Grounding , author=. arXiv preprint arXiv:2606.14579 , year=

  25. [33]

    arXiv preprint arXiv:2509.21552 , year=

    Learning gui grounding with spatial reasoning from visual feedback , author=. arXiv preprint arXiv:2509.21552 , year=

    arXiv

  26. [34]

    arXiv preprint arXiv:2511.07332 , year=

    Grounding Computer Use Agents on Human Demonstrations , author=. arXiv preprint arXiv:2511.07332 , year=

    arXiv

  27. [35]

    Gemini 3.1 Pro Preview , year =

  28. [36]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Cutpaste: Self-supervised learning for anomaly detection and localization , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  29. [37]

    Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=

    Measure Twice, Click Once: Co-evolving Proposer and Visual Critic via Reinforcement Learning for GUI Grounding , author=. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=

  30. [38]

    2025 IEEE/CVF International Conference on Computer Vision (ICCV) , pages=

    Falcon: Resolving visual redundancy and fragmentation in high-resolution multimodal large language models via visual registers , author=. 2025 IEEE/CVF International Conference on Computer Vision (ICCV) , pages=. 2025 , organization=

    2025

  31. [39]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    Zhou, Xurui and Chen, Gongwei and Xie, Yuquan and Li, Zaijing and Zhou, Kaiwen and Wang, Shuai and Yang, Shuo and Tian, Zhuotao and Shao, Rui , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

  32. [40]

    Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=

    Personalalign: Hierarchical implicit intent alignment for personalized gui agent with long-term user-centric records , author=. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=

  33. [41]

    2025 IEEE/CVF International Conference on Computer Vision (ICCV) , pages=

    Less is more: Empowering gui agent with context-aware simplification , author=. 2025 IEEE/CVF International Conference on Computer Vision (ICCV) , pages=. 2025 , organization=

    2025

  34. [42]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

    Detecting and grounding multi-modal media manipulation and beyond , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

  35. [43]

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

    Detecting and grounding multi-modal media manipulation , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  36. [44]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Multi-adversarial discriminative deep domain generalization for face presentation attack detection , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  37. [45]

    Advances in neural information processing systems , volume=

    CogVLA: Cognition-aligned vision-language-action models via instruction-driven routing & sparsification , author=. Advances in neural information processing systems , volume=

  38. [46]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Semanticvla: Semantic-aligned sparsification and enhancement for efficient robotic manipulation , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  39. [47]

    International Conference on Learning Representations , year=

    From Abstraction to Instantiation: Learning Behavioral Representation for Vision-Language-Action Model , author=. International Conference on Learning Representations , year=

  40. [48]

    2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

    Lion-fs: Fast & slow video-language thinker as online video assistant , author=. 2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=. 2025 , organization=

    2025

  41. [49]

    arXiv preprint arXiv:2506.10357 , year=

    Optimus-3: Dual-router aligned mixture-of-experts agent with dual-granularity reasoning-aware policy optimization , author=. arXiv preprint arXiv:2506.10357 , year=

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.