Pith. sign in

REVIEW 3 major objections 5 minor 83 references

Test-Time Self-Evolving GUI Visual Grounding via Reflection-Guided On-Policy Self-Distillation

T0 review · 3 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a GUI grounding model can keep improving after deployment, with no human labels, by reflecting on its own click predictions and distilling that reflection into dense per-token training feedback.

desk verdict A genuine new combination for test-time GUI grounding, with broad experiments and a thoughtful contrastive calibration mechanism; the main gap is that the reflection text's content is never directly verified. read the letter →

arxiv 2608.11191 v1 pith:CA3RCSDO submitted 2026-08-11 cs.CV cs.AIcs.CL

Shiyu Xuan , Zechao Li This is my paper · ORCID
classification cs.CVcs.AIcs.CL
keywords GUIvisualgroundingtest-timeadaptationon-policyself-distillationcontrastivecalibrationmultimodallanguagemodelreflectionself-evolvingagents
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper sets out to close the post-deployment adaptation gap for GUI visual grounding: models that localize UI elements from natural-language instructions typically freeze after training, so they never get better on new apps or layouts. The authors propose a self-evolving loop in which the model explores an unseen interface, an MLLM-based Reflector scores whether each predicted coordinate satisfies the instruction and writes a short reasoning about why, and a Reflection-Guided On-Policy Self-Distillation step converts that reasoning into per-token supervision. A Contrastive Calibration trick prevents the auto-regressive generation of wrong coordinate tokens from poisoning the training signal. Across six benchmarks the framework reports an average accuracy gain of 7.4 points over the frozen base model and up to 7.7 points over a recent test-time reinforcement learning baseline, suggesting that reflection-grounded self-distillation could let deployed GUI agents learn from their own experience.

What carries the argument

The central object is Reflection-Guided On-Policy Self-Distillation (R-OPSD) combined with Contrastive Calibration. R-OPSD builds a self-teacher by conditioning the same grounding model on the privileged information (the Reflector's binary score S and free-text reflection R), and defines the token-level advantage as the log-probability ratio between this teacher and the student's own policy on the student's generated coordinate tokens. Contrastive Calibration replaces the student in the denominator with an inverse-prompted student (which is told the failed prediction was correct), so that at the first incorrect token the teacher assigns lower probability to the wrong token than the deluded student, producing a negative advantage, while on drifted prefixes the two distributions converge and the advantage shrinks toward zero, preventing corrupted feedback.

What would settle it

On a held-out set of failed rollouts, measure the average absolute token-level advantage on the final drifted tokens: the paper reports it decays to 0.0033 with Contrastive Calibration, so a replication finding values above 0.1 would refute the claim that corrupted prefixes are filtered. A second check is to feed the Reflector's binary verdicts with shuffled text reflections; if gains persist, the reasoning text is not the load-bearing signal.

Watch

Extended reading notes

Core claim

The central claim is that textual reflection on a failed or successful click, produced by a frozen multimodal Reflector, can be internalized into a GUI grounding model's weights through on-policy self-distillation, turning high-level natural-language reasoning into dense token-level advantages. The paper argues that the existing test-time RL baseline relies on sparse scalar rewards and therefore cannot explain or correct failures, whereas the reflection-conditioned self-teacher provides a superior gradient signal even on fully failed rollout groups that defeat relative-reward methods like GRPO. To keep that signal clean, the Contrastive Calibration method computes the advantage against an inverse-prompted student that treats the failed rollout as correct, yielding a negative penalty at the first wrong token and an advantage that decays to near zero once the prefix has drifted. The paper's evidence is the consistent gains across six benchmarks and two base model families, plus ablations showing that removing reflection, calibration, or direction clamping causes collapse or negative transfer.

Load-bearing premise

The whole loop works only if the Reflector's binary verdict and free-text reasoning are informative enough to improve the teacher's conditional distribution; the paper also assumes, without proof, that on drifted prefixes the inverse-prompted student and the reflection-guided teacher converge so that corrupted advantages decay to zero.

Editorial extensions

If this is right

  • Deployed GUI agents can adapt to unseen applications and layouts without any human-annotated ground truth, learning from both successes and failures.
  • Failed explorations, which are typically discarded or merely penalized by sparse rewards, are converted into dense per-token training signal.
  • The framework consistently improves accuracy across six GUI benchmarks and on both 2B/3B and 7B/8B base models, and outperforms the test-time RL baseline GUI-RCPO.
  • The shared base model with alternating LoRA adapters keeps the memory footprint low (about 10 GB for 3B/2B models, 30 GB for 7B/8B models), making self-evolution feasible on a single GPU.
  • R-OPSD without the query-level GRPO advantage already runs at roughly one-third the training time of standard GRPO, so the core distillation loop is the cheap and effective part.

Reading between the lines

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

  • The same Exploration-Evaluation-Reflection-Internalization loop could transfer to other coordinate-output tasks, such as robot manipulation or document layout, whenever a verifier can produce textual diagnostics.
  • The framework's ceiling is set by the Reflector's quality; a natural stress test is to measure how the grounding gains scale with Reflector accuracy, or to inject deliberately misleading reflections while keeping the binary score fixed.
  • Contrastive calibration is a general cure for drifted auto-regressive prefixes and could be applied to other on-policy distillation settings beyond GUI grounding, such as long-horizon agent trajectories.
  • The approach implicitly assumes the reflection text, not just the binary verdict, carries information; an ablation that shuffles or empties the reflection while keeping S would isolate where the gain actually comes from.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes a test-time self-evolving framework for GUI visual grounding. A frozen grounding policy performs rollouts on unlabeled screenshots; an MLLM-based Reflector evaluates each predicted coordinate and outputs a binary score S together with a textual reflection R; then Reflection-Guided On-Policy Self-Distillation (R-OPSD) converts S and R into token-level advantages by comparing a reflection-conditioned self-teacher with the base policy. A Contrastive Calibration method, direction-based advantage clamping, and an optional query-level GRPO term are added. Experiments on ScreenSpot-v2, ScreenSpot-Pro, OSWorld-G, OSWorld-G-Refine, and MMBench-GUI report an average improvement of 7.4% over the base model and favorable comparisons with GUI-RCPO. The appendix includes Reflector accuracy, scalability to 7B/8B models, computational overhead, and an unsupervised training setting.

Significance. If the reported results are robust, the paper would be a meaningful step toward post-deployment adaptation of GUI grounding models without human-labeled ground truth. The framework combines several established ideas in a new configuration: an MLLM Reflector, on-policy self-distillation, and a contrastive mechanism for avoiding corrupted prefix supervision. The experimental coverage is broad, with six benchmarks, two backbone families, component ablations, a separate Reflector reliability study, and an overhead analysis. The main technical elements are clearly described and the design is plausible. The credibility of the central claim, however, rests on two points that are currently under-supported: whether the reflection text R is genuinely informative, and whether the empirical gains are statistically reliable rather than artifacts of single runs and benchmark-level hyperparameter selection.

major comments (3)
  1. [Evaluation and Reflection with Reflector; Eq. (2); Table 4] The central claim that the model learns from the Reflector's diagnostic reasoning is not directly supported, because the quality of the reflection R is never measured. Eq. (2) provides reward only for the binary label S and the output format; the reflection text receives no direct supervision or evaluation metric. Table 4 reports S accuracy, precision, and recall, but no metric for R. The ablation that adds or removes the reflection field (Table 2, rows (d)-(e) and (i)-(j)) changes the teacher prompt as a whole and therefore cannot separate the content of R from the mere presence of an additional text field. Please add a corruption or sensitivity analysis of R (for example, replacing R with templated or generic reasoning, truncating or shuffling the reflection, or randomly swapping reflections between samples) and report the resulting grounding accuracy. Alternatively, provide a direct measure of R informativeness on a held-out set with human annotations. Without this, the statement in the Methodology discussion that "R-OPSD only assumes that the Reflector provides sufficiently informative evaluations and reflections" remains a stated assumption, and the attribution of the 7.4% gain to reflection-guided supervision is not established.
  2. [Experiments; Tables 1-4; Fig. 4] The headline result is not statistically supported as reported. Every accuracy in Tables 1, 2, 3, and 6 is a single run without standard deviations, confidence intervals, or significance tests, and Fig. 4 reports averages without variance or sample size. Some of the claimed gains are small relative to typical benchmark noise; for example, Table 1 shows 61.2% vs. 60.8% on OSWorld-G-Refine for Qwen3-VL-2B adapted on SSv2. In addition, the integration strength lambda is selected using the evaluation benchmarks themselves in Fig. 5, and the CC and clamping components are ablated and selected on the same benchmarks in Table 2. Please report multi-seed intervals and significance tests, state the number of rollouts used in Fig. 4, and describe a hyperparameter-selection procedure that does not use target benchmark labels. These additions are necessary to support the average improvement claim as a robust effect rather than a tuned point estimate.
  3. [Implementation Details; Table 1] The evaluation protocol should be clarified with respect to data provenance. The paper says the model adapts on "data from ScreenSpot-v2 or MMBench-GUI, strictly omitting their ground-truth annotations," and then reports accuracy on the same datasets. Please specify whether the unlabeled adaptation samples are drawn from the test splits of those benchmarks, whether the frozen baselines are evaluated under identical inference conditions, and how the adaptation procedure avoids using evaluation labels for early stopping, model selection, or lambda selection. The current description does not distinguish transductive test-time adaptation from unlabeled domain adaptation, and this distinction is relevant to the "unseen interfaces" claim in the abstract and to the fairness of the comparison with GUI-RCPO.
minor comments (5)
  1. [Introduction] The phrase "This is the first work to applying OPSD in test-time scenarios" should be corrected to "the first work to apply OPSD."
  2. [Ablation Study; Table 2] The formatting of Table 2 makes the exact configuration of each row difficult to parse, especially because collapsed rows (a) and (f) are marked with "×" in both the component columns and the result columns. Please separate the component settings from the numeric results or spell out the configuration of every row in the caption.
  3. [Appendix; Table 5] The statement that R-OPSD with CC "only takes 151 minutes, reducing the training time by more than 34% compared to GRPO" applies to the configuration without the query-level advantage; the full framework in row "R-OPSD (w/ CC & QA)" actually costs 1.73x of GRPO. This distinction should be stated in the main text so readers do not infer that the final method is cheaper than the GRPO baseline.
  4. [Fig. 4] Figure 4 would be more informative if it reported the number of failed rollouts used for the averages and showed per-token distributions or error bars; as presented, the two bars are aggregate evidence for a mechanism that is claimed to operate at the level of individual drift steps.
  5. [Discussion of R-OPSD, Eq. (5)] Because the teacher and the student share the same parameters, please state explicitly in the text that the log-probabilities in the advantage are evaluated under a frozen or stop-gradient snapshot of pi_G before the update, so that the objective is a distribution-matching loss rather than an unrolled gradient through the teacher.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central improvement claim is verified against external ground-truth benchmarks, and the Reflector and R-OPSD components are empirically evaluated rather than defined into existence.

full rationale

The paper's central claim—an average accuracy improvement of 7.4% over the base model—is a measured outcome on six external GUI grounding benchmarks (Table 1), not a quantity constructed from the method's own definitions. The Reflector is trained offline on GroundCUA with binary labels derived from ground-truth bounding boxes (Eq. 2), and its evaluation accuracy is separately reported on a held-out set (Table 4); this gives the evaluation signal independent empirical content. The R-OPSD advantage (Eq. 5) compares the same policy's likelihood under different conditioning prompts, which is self-distillation by design rather than circular derivation: the claim is that this objective improves accuracy, and Table 2 and Table 3 test that claim against external labels. The paper explicitly states its assumption that 'the Reflector provides sufficiently informative evaluations and reflections' rather than deriving that as a theorem, so the assumption is a limitation, not a circular step. The adaptation on ScreenSpot-v2 or MMBench-GUI and evaluation on those same datasets is consistent with the test-time adaptation setting; additional benchmarks (SS, SSP, OSW-G, OSW-GR) provide out-of-distribution checks. The only self-citation (MMBench-GUI, which includes an author of this paper) is a benchmark reference and is not load-bearing. No equation reduces to its own input or renames a fitted parameter as a prediction.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the assumption that the Reflector provides useful feedback and that the contrastive calibration's convergence behavior holds; these are stated but not formally proven. The main tuned hyperparameter is lambda=0.2, selected on the evaluation benchmarks.

free parameters (3)
  • lambda (integration strength) = 0.2
    Chosen via sensitivity analysis on SSv2 and MMG (Fig. 5); the final reported results use this value, so the headline gains are partly tuned on the evaluation benchmarks.
  • rollout sampling temperature = 1.0
    Set by hand for exploration during rollout; affects the distribution of explored trajectories and thus the feedback quality.
  • top-p = 0.95
    Set by hand for nucleus sampling during rollout; standard choice but affects exploration diversity.
assumptions (4)
  • standard math The policy gradient / OPD objective with token-level advantages is a valid optimization target.
    Uses log-ratio advantages in Eqs. 3-5; standard in the RL and on-policy distillation literature.
  • domain assumption GUI grounding can be cast as auto-regressive coordinate token generation.
    The methodology section and experiments treat coordinate outputs as generated tokens, which is required for token-level advantages.
  • ad hoc to paper The Reflector provides sufficiently informative evaluations and reflections to improve the teacher's conditional distribution.
    Explicitly stated in the Internalization section: 'R-OPSD only assumes that the Reflector provides sufficiently informative evaluations and reflections to improve the conditional distribution of the teacher.'
  • ad hoc to paper The teacher and inverse-prompted student distributions converge on drifted prefixes, so the contrastive advantage decays to nearly zero.
    Contrastive Calibration Method section: 'As more incorrect tokens are generated, both models are conditioned on increasingly similar prefixes, leading to similar probabilities ... driving the advantage to almost zero.' This is a heuristic claim supported only by aggregate statistics.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Test-Time Self-Evolving GUI Visual Grounding via Reflection-Guided On-Policy Self-Distillation." pith.science (2026). https://pith.science/paper/CA3RCSDO

@misc{pith2026260811191,
  author       = {Pith},
  title        = {Pith review of: Test-Time Self-Evolving GUI Visual Grounding via Reflection-Guided On-Policy Self-Distillation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CA3RCSDO}},
  note         = {Machine review of arXiv:2608.11191}
}
read the original abstract

GUI Visual Grounding is a fundamental capability for GUI agents. Existing models typically freeze their parameters after deployment, limiting their ability to adapt to unseen interfaces. Although recent methods attempt to adapt models via test-time reinforcement learning, they cannot reflect upon failed exploration. To overcome this, we propose a Test-Time Self-Evolving framework that enables models to improve after deployment without human-annotated ground truth. It constructs a closed-loop of Exploration, Evaluation, Reflection, and Internalization. Specifically, the agent first explores unseen interfaces by predicting grounding coordinates for given instructions. To evaluate these explorations, we introduce an MLLM-based Reflector to assess the generated results and provide the corresponding reasoning reflections. To internalize reflection knowledge into the model weights, we propose Reflection-Guided On-Policy Self-Distillation, which translates high-level reasoning into dense token-level supervision via a conditioned self-teacher. Furthermore, we design a Contrastive Calibration method to prevent incorrect auto-regressive prefixes from corrupting the supervisory signals during failed explorations. Extensive experiments across six benchmarks demonstrate our framework's effectiveness, achieving an average accuracy improvement of 7.4% over the base model. To the best of our knowledge, this is the first work to successfully exploit on-policy self-distillation for test-time adaptation in GUI visual grounding. By filling the gap in post-deployment adaptation, our framework completes the self-evolving capability of GUI agents. The code will be released.

Figures

Figures reproduced from arXiv: 2608.11191 by the authors.

Figure 1
Figure 1. (a) Static methods operate on frozen parameters af [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Test-Time Self-Evolving Framework consists of four core stages: Exploration, Evaluation, Reflection, and Internal￾ization. During test-time adaptation, πG explores the environment to predict coordinates. The Reflector πR then assesses the prediction to generate a binary score S and a step-by-step reasoning R. The reflection knowledge is internalized via Reflection￾Guided On-Policy Self-Distillation to update the gro… view at source ↗
Figure 4
Figure 4. Quantitative analysis of token-level advantage on [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figures from the paper (3 more)
Figure 3
Figure 3. Figure 3: Illustration of Contrastive Calibration Method. Dur [PITH_FULL_IMAGE:figures/full_fig_p005_3.png]
Figure 5
Figure 5. Figure 5: Sensitivity analysis of the integration strength [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Qualitative visualizations of the Reflector’s evaluation and reasoning. The red bounding boxes denote the predicted [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

83 extracted references · 18 canonical work pages

  1. [1]

    Communication, Simulation, and Intelligent Agents: Implications of Personal Intelligent Machines for Medical Education

    Clancey, William J. Communication, Simulation, and Intelligent Agents: Implications of Personal Intelligent Machines for Medical Education. Proceedings of the Eighth International Joint Conference on Artificial Intelligence (IJCAI-83)

  2. [2]

    Classification Problem Solving

    Clancey, William J. Classification Problem Solving. Proceedings of the Fourth National Conference on Artificial Intelligence

  3. [3]

    , title =

    Robinson, Arthur L. , title =. 1980 , doi =. https://science.sciencemag.org/content/208/4447/1019.full.pdf , journal =

    1980

  4. [4]

    New Ways to Make Microcircuits Smaller---Duplicate Entry

    Robinson, Arthur L. New Ways to Make Microcircuits Smaller---Duplicate Entry. Science

  5. [5]

    Clancey and Glenn Rennels , abstract =

    Diane Warner Hasling and William J. Clancey and Glenn Rennels , abstract =. Strategic explanations for a diagnostic consultation system , journal =. 1984 , issn =. doi:https://doi.org/10.1016/S0020-7373(84)80003-6 , url =

    doi:10.1016/s0020-7373(84)80003-6 1984

  6. [6]

    and Rennels, Glenn R

    Hasling, Diane Warner and Clancey, William J. and Rennels, Glenn R. and Test, Thomas. Strategic Explanations in Consultation---Duplicate. The International Journal of Man-Machine Studies

  7. [7]

    Poligon: A System for Parallel Problem Solving

    Rice, James. Poligon: A System for Parallel Problem Solving

  8. [8]

    Transfer of Rule-Based Expertise through a Tutorial Dialogue

    Clancey, William J. Transfer of Rule-Based Expertise through a Tutorial Dialogue

Show all 83 references
  1. [9]

    The Engineering of Qualitative Models

    Clancey, William J. The Engineering of Qualitative Models

  2. [10]

    2023 , eprint=

    Attention Is All You Need , author=. 2023 , eprint=

    2023

  3. [11]

    Pluto: The 'Other' Red Planet

    NASA. Pluto: The 'Other' Red Planet

  4. [12]

    arXiv preprint arXiv:2604.03128 , year=

    Self-Distilled RLVR , author=. arXiv preprint arXiv:2604.03128 , year=

    arXiv

  5. [13]

    arXiv preprint arXiv:1503.02531 , year=

    Distilling the knowledge in a neural network , author=. arXiv preprint arXiv:1503.02531 , year=

    arXiv

  6. [14]

    arXiv preprint arXiv:2602.12687 , year=

    Trust the uncertain teacher: distilling dark knowledge via calibrated uncertainty , author=. arXiv preprint arXiv:2602.12687 , year=

    arXiv

  7. [15]

    arXiv preprint arXiv:2602.02244 , year=

    Learning While Staying Curious: Entropy-Preserving Supervised Fine-Tuning via Adaptive Self-Distillation for Large Reasoning Models , author=. arXiv preprint arXiv:2602.02244 , year=

    arXiv

  8. [16]

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

    Todi: Token-wise distillation via fine-grained divergence control , author=. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing , pages=

    2025

  9. [17]

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

    EA-KD: Entropy-based adaptive knowledge distillation , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  10. [18]

    arXiv preprint arXiv:2505.15810 , year=

    Gui-g1: Understanding r1-zero-like training for visual grounding in gui agents , author=. arXiv preprint arXiv:2505.15810 , year=

    arXiv

  11. [19]

    AAAI , volume=

    GUI-G ^2 : Gaussian Reward Modeling for GUI Grounding , author=. AAAI , volume=

  12. [20]

    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

  13. [21]

    arXiv preprint arXiv:2505.12370 , year=

    Enhancing visual grounding for gui agents via self-evolutionary reinforcement learning , author=. arXiv preprint arXiv:2505.12370 , year=

    arXiv

  14. [22]

    ACM MM , pages=

    Screenspot-pro: Gui grounding for professional high-resolution computer use , author=. ACM MM , pages=

  15. [23]

    arXiv preprint arXiv:2505.13227 , year=

    Scaling computer-use grounding via user interface decomposition and synthesis , author=. arXiv preprint arXiv:2505.13227 , year=

  16. [24]

    arXiv preprint arXiv:2503.15661 , year=

    Ui-vision: A desktop-centric gui benchmark for visual perception and interaction , author=. arXiv preprint arXiv:2503.15661 , year=

    arXiv

  17. [25]

    arXiv preprint arXiv:2507.19478 , year=

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

    arXiv

  18. [26]

    Os-atlas: A foundation action model for generalist gui agents, 2024 , author=

    2024

  19. [27]

    arXiv preprint arXiv:2511.21631 , year=

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

    arXiv

  20. [28]

    arXiv preprint arXiv:2509.15221 , year=

    Scalecua: Scaling open-source computer use agents with cross-platform data , author=. arXiv preprint arXiv:2509.15221 , year=

  21. [29]

    arXiv preprint arXiv:2501.12326 , year=

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

    arXiv

  22. [30]

    arXiv preprint arXiv:2603.14448 , year=

    Zoom to Essence: Trainless GUI Grounding by Inferring upon Interface Elements , author=. arXiv preprint arXiv:2603.14448 , year=

  23. [31]

    arXiv preprint arXiv:2510.20286 , year=

    UI-Ins: Enhancing GUI Grounding with Multi-Perspective Instruction-as-Reasoning , author=. arXiv preprint arXiv:2510.20286 , year=

  24. [32]

    arXiv preprint arXiv:2510.27266 , year=

    HyperClick: Advancing Reliable GUI Grounding via Uncertainty Calibration , author=. arXiv preprint arXiv:2510.27266 , year=

    arXiv

  25. [33]

    arXiv preprint arXiv:2506.09373 , year=

    Lpo: Towards accurate gui agent interaction via location preference optimization , author=. arXiv preprint arXiv:2506.09373 , year=

    arXiv

  26. [34]

    arXiv e-prints , pages=

    Tongui: Building generalized gui agents by learning from multimodal web tutorials , author=. arXiv e-prints , pages=

  27. [35]

    arXiv preprint arXiv:2506.03143 , year=

    Gui-actor: Coordinate-free visual grounding for gui agents , author=. arXiv preprint arXiv:2506.03143 , year=

    arXiv

  28. [36]

    arXiv preprint arXiv:2507.05791 , year=

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

    arXiv

  29. [37]

    arXiv preprint arXiv:2602.11858 , year=

    Zooming without Zooming: Region-to-Image Distillation for Fine-Grained Multimodal Perception , author=. arXiv preprint arXiv:2602.11858 , year=

  30. [38]

    arXiv preprint arXiv:2604.08516 , year=

    MolmoWeb: Open Visual Web Agent and Open Data for the Open Web , author=. arXiv preprint arXiv:2604.08516 , year=

    arXiv

  31. [39]

    arXiv preprint arXiv:2604.21268 , year=

    Measure Twice, Click Once: Co-evolving Proposer and Visual Critic via Reinforcement Learning for GUI Grounding , author=. arXiv preprint arXiv:2604.21268 , year=

    arXiv

  32. [40]

    arXiv preprint arXiv:2410.05243 , year=

    Navigating the digital world as humans do: Universal visual grounding for gui agents , author=. arXiv preprint arXiv:2410.05243 , year=

    arXiv

  33. [41]

    ACL , pages=

    Seeclick: Harnessing gui grounding for advanced visual gui agents , author=. ACL , pages=

  34. [42]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Infigui-g1: Advancing gui grounding with adaptive exploration policy optimization , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  35. [43]

    arXiv preprint arXiv:2601.09770 , year=

    GUI-Eyes: Tool-Augmented Perception for Visual Grounding in GUI Agents , author=. arXiv preprint arXiv:2601.09770 , year=

  36. [44]

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

    Gui-bee: Align gui action grounding to novel environments via autonomous exploration , author=. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing , pages=

    2025

  37. [45]

    arXiv preprint arXiv:2402.03300 , year=

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

    arXiv

  38. [46]

    arXiv preprint arXiv:2501.12948 , year=

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

    arXiv

  39. [47]

    arXiv preprint arXiv:2601.18734 , year=

    Self-Distilled Reasoner: On-Policy Self-Distillation for Large Language Models , author=. arXiv preprint arXiv:2601.18734 , year=

    arXiv

  40. [48]

    arXiv preprint arXiv:2601.19897 , year=

    Self-Distillation Enables Continual Learning , author=. arXiv preprint arXiv:2601.19897 , year=

    arXiv

  41. [49]

    arXiv preprint arXiv:2601.18779 , year=

    POPE: Learning to Reason on Hard Problems via Privileged On-Policy Exploration , author=. arXiv preprint arXiv:2601.18779 , year=

  42. [50]

    arXiv preprint arXiv:2601.20802 , year=

    Reinforcement Learning via Self-Distillation , author=. arXiv preprint arXiv:2601.20802 , year=

    arXiv

  43. [51]

    arXiv preprint arXiv:2602.02482 , year=

    Expanding the Capabilities of Reinforcement Learning via Text Feedback , author=. arXiv preprint arXiv:2602.02482 , year=

  44. [52]

    arXiv preprint arXiv:2604.00626 , year=

    A Survey of On-Policy Distillation for Large Language Models , author=. arXiv preprint arXiv:2604.00626 , year=

    arXiv

  45. [53]

    arXiv preprint arXiv:2604.17535 , year=

    OPSDL: On-Policy Self-Distillation for Long-Context Language Models , author=. arXiv preprint arXiv:2604.17535 , year=

    arXiv

  46. [54]

    arXiv preprint arXiv:2602.09082 , year=

    UI-Venus-1.5 Technical Report , author=. arXiv preprint arXiv:2602.09082 , year=

  47. [55]

    arXiv preprint arXiv:2509.18119 , year=

    Mobilerl: Online agentic reinforcement learning for mobile gui agents , author=. arXiv preprint arXiv:2509.18119 , year=

  48. [56]

    arXiv preprint arXiv:2508.14040 , year=

    Computerrl: Scaling end-to-end online reinforcement learning for computer use agents , author=. arXiv preprint arXiv:2508.14040 , year=

  49. [57]

    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

  50. [58]

    AAAI , volume=

    Ui-r1: Enhancing efficient action prediction of gui agents by reinforcement learning , author=. AAAI , volume=

  51. [59]

    arXiv preprint arXiv:2604.13016 , year=

    Rethinking On-Policy Distillation of Large Language Models: Phenomenology, Mechanism, and Recipe , author=. arXiv preprint arXiv:2604.13016 , year=

    arXiv

  52. [60]

    arXiv preprint arXiv:2508.04389 , year=

    GuirlVG: Incentivize GUI Visual Grounding via Empirical Exploration on Reinforcement Learning , author=. arXiv preprint arXiv:2508.04389 , year=

    arXiv

  53. [61]

    arXiv preprint arXiv:2603.05044 , year=

    WebFactory: Automated Compression of Foundational Language Intelligence into Grounded Web Agents , author=. arXiv preprint arXiv:2603.05044 , year=

    arXiv

  54. [62]

    arXiv preprint arXiv:2505.12299 , year=

    MobileIPL: Enhancing Mobile Agents Thinking Process via Iterative Preference Learning , author=. arXiv preprint arXiv:2505.12299 , year=

  55. [63]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    FDC-Ground: Improving GRPO for GUI Grounding via Exponential Rewards and Fact-Aligned Pruning , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  56. [64]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Co-EPG: A Framework for Co-Evolution of Planning and Grounding in Autonomous GUI Agents , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  57. [65]

    arXiv preprint arXiv:2509.15566 , year=

    Btl-ui: Blink-think-link reasoning model for gui agent , author=. arXiv preprint arXiv:2509.15566 , year=

  58. [66]

    CVPR , pages=

    Showui: One vision-language-action model for gui visual agent , author=. CVPR , pages=

  59. [67]

    arXiv preprint arXiv:2601.03928 , year=

    FocusUI: Efficient UI Grounding via Position-Preserving Visual Token Selection , author=. arXiv preprint arXiv:2601.03928 , year=

  60. [68]

    ICCV , pages=

    Visual test-time scaling for gui agent grounding , author=. ICCV , pages=

  61. [69]

    AAAI , volume=

    Test-time reinforcement learning for gui grounding via region consistency , author=. AAAI , volume=

  62. [70]

    NeurIPS , volume=

    Ttrl: Test-time reinforcement learning , author=. NeurIPS , volume=

  63. [71]

    ICLR , volume=

    Minillm: Knowledge distillation of large language models , author=. ICLR , volume=

  64. [72]

    ICLR , volume=

    On-policy distillation of language models: Learning from self-generated mistakes , author=. ICLR , volume=

  65. [73]

    arXiv preprint arXiv:2605.00642 , year=

    Learn where to Click from Yourself: On-Policy Self-Distillation for GUI Grounding , author=. arXiv preprint arXiv:2605.00642 , year=

    arXiv

  66. [74]

    arXiv preprint arXiv:2605.27028 , year=

    Less is More: Early Stopping Rollout for On-Policy Distillation , author=. arXiv preprint arXiv:2605.27028 , year=

    arXiv

  67. [75]

    arXiv preprint arXiv:2603.07079 , year=

    Entropy-Aware On-Policy Distillation of Language Models , author=. arXiv preprint arXiv:2603.07079 , year=

    arXiv

  68. [76]

    arXiv preprint arXiv:2602.02994 , year=

    Video-OPD: Efficient Post-Training of Multimodal Large Language Models for Temporal Video Grounding via On-Policy Distillation , author=. arXiv preprint arXiv:2602.02994 , year=

    arXiv

  69. [77]

    arXiv preprint arXiv:2605.18740 , year=

    Vision-opd: Learning to see fine details for multimodal llms via on-policy self-distillation , author=. arXiv preprint arXiv:2605.18740 , year=

    arXiv

  70. [78]

    Frontiers of Computer Science , volume=

    A survey on large language model based autonomous agents , author=. Frontiers of Computer Science , volume=. 2024 , publisher=

    2024

  71. [79]

    arXiv preprint arXiv:2604.02288 , year=

    Unifying group-relative and self-distillation policy optimization via sample routing , author=. arXiv preprint arXiv:2604.02288 , year=

  72. [80]

    arXiv preprint arXiv:2606.11709 , year=

    RLCSD: Reinforcement Learning with Contrastive On-Policy Self-Distillation , author=. arXiv preprint arXiv:2606.11709 , year=

    arXiv

  73. [81]

    , author=

    Lora: Low-rank adaptation of large language models. , author=. ICLR , volume=

  74. [82]

    arXiv preprint arXiv:2511.07332 , year=

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

    arXiv

  75. [83]

    Qwen2.5-VL , url =

    Qwen Team , month =. Qwen2.5-VL , url =

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.