Pith. sign in

REVIEW 4 major objections 5 minor 47 references

ViDDAR: Vision Language Model-Based Task-Detrimental Content Detection for Augmented Reality

T0 review · 4 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read ViDDAR claims that a cloud vision-language model plus edge segmentation can detect AR content that hides or misrepresents real-world objects, reaching 92.15% obstruction accuracy and 82.46% manipulation accuracy.

desk verdict A credible VLM-based AR safety detector with a useful dataset, but the reported accuracy is narrower than the formal model because it only tests single-key-object scenes and the threshold is fit to the data. read the letter →

arxiv 2501.12553 v2 pith:OU334ZYP submitted 2025-01-22 cs.CV

Yanming Xiu , Tim Scargill , Maria Gorlatova This is my paper · ORCID
classification cs.CV
keywords AugmentedrealityVisionlanguagemodelsObstructionattackInformationmanipulationARcontentsafetyObjectdetectionSegmentationEdge-cloudlatency
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that a vision language model running in the cloud can act as an augmented reality safety monitor, catching two kinds of harmful virtual overlays: obstruction, where virtual content covers an important real object, and information manipulation, where virtual content makes a user misread what a real object is or does. The authors propose ViDDAR, a full-reference system that compares the raw camera frame with the augmented frame, uses a VLM to decide which real object in a scene matters, then checks whether virtual content overlaps it enough to count as an obstruction or misleads the user semantically. On a self-collected dataset of 306 obstruction image pairs and 114 manipulation image pairs, the system reports up to 92.15% obstruction detection accuracy with 533 ms latency and 82.46% manipulation detection accuracy with 9.62 s latency when using GPT-4o. The paper matters because it demonstrates a path beyond pixel-level or saliency-based AR content assessment toward semantic, context-dependent safety checks that adapt to what the user is actually looking at.

What carries the argument

The load-bearing mechanism is the mathematical definition of an obstruction attack, O(Ir, Ia) = 1 if any key-object mask overlaps the virtual-content mask by at least alpha of the object area, paired with the three-factor Boolean model M = A AND S AND I for information manipulation. The implemented pipeline realizing these definitions is a user-edge-cloud architecture: the AR device captures raw and augmented frames; the edge server runs Grounding DINO to detect the VLM-named key object and SAM to produce its binary mask, with the virtual-content mask obtained by pixel-level comparison of the two frames; and the cloud hosts a VLM, either GPT-4o or LLaVA-Next-8b, that identifies the key object or scores the manipulation factors from a crafted prompt. This decomposition matters because it turns a semantic judgment, namely which object matters in this scene, into a mechanically checkable overlap ratio, and it turns a subjective manipulation judgment into a checklist of perceptually grounded Boolean conditions.

What would settle it

Run the full ViDDAR obstruction pipeline on new image pairs containing several competing important objects, such as a pedestrian, a stop sign, and a warning cone in one frame, and measure key-object recognition accuracy; if it drops materially below the reported 91.83%, or if a sweep across obstruction overlap ratios shows the alpha = 0.25 boundary misclassifying partial occlusions, the claim that the system generalizes beyond its dataset is falsified.

Watch

Extended reading notes

Core claim

The central claim is that task-detrimental AR content can be separated into two formally defined attack types and detected by a VLM-driven full-reference pipeline. An obstruction attack is defined as existing when, for at least one key object, the pixel overlap between the key object mask and the virtual content mask reaches or exceeds a fraction alpha of the key object's mask; the system's task therefore becomes dynamic key-object recognition plus segmentation plus mask comparison. An information manipulation attack is defined as the conjunction of three Boolean conditions -- alignment precision, style similarity, and information misrepresentation -- and ViDDAR asks a VLM six targeted questions to evaluate them. With GPT-4o the system reaches 92.15% obstruction accuracy, close to the 93.14% upper bound obtained by supplying the key object in advance, and 82.46% manipulation accuracy, while traditional saliency and edge baselines for obstruction stay near chance. ViDDAR is presented as the first system to employ VLMs for detecting task-detrimental content in AR settings.

Load-bearing premise

The pipeline assumes the cloud VLM's single key-object choice, validated on 306 images of 23 object classes, transfers to arbitrary AR scenes, and that the 0.25 obstruction threshold calibrated to this dataset holds elsewhere.

Editorial extensions

If this is right

  • If ViDDAR is correct, AR devices could automatically reduce the opacity of virtual content when it covers a safety-critical object, without requiring developers to predefine which objects are important in every scene.
  • Obstruction detection runs at interactive latency, 533 ms in a one-hop network, making it feasible as a per-frame or near-per-frame safety check in navigation and task-guidance applications.
  • Information manipulation detection is much slower, 9.62 s, so it is suitable for periodic checks of stable content rather than frame-by-frame evaluation.
  • Traditional saliency and edge-based methods perform near chance on obstruction detection, roughly 52%, which supports the claim that semantic understanding is necessary for this task.
  • The small accuracy gap between ViDDAR and the prior-knowledge upper bound, about one percentage point, indicates that key-object recognition by the VLM, not segmentation or mask comparison, is the main bottleneck for obstruction detection.

Reading between the lines

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

  • Beyond the paper, the 0.25 obstruction threshold is read off this dataset's boundary, so real deployment would need to measure how detection accuracy varies with alpha on scenes with partial occlusions, where safety decisions may require a more conservative threshold.
  • Beyond the paper, the single-key-object prompt is a deliberate simplification; in scenes with several competing important objects, a priority-ranked list of key objects would likely outperform both the single-object choice and the greedy all-object strategy that the paper shows produces false positives.
  • Beyond the paper, the AND logic for information manipulation treats alignment, style similarity, and misrepresentation as independent and necessary conditions; a testable extension is whether human perception of manipulation is better modeled by a graded or compensatory combination, for example very strong misrepresentation outweighing imperfect alignment.
  • Beyond the paper, the architecture implies that a smaller VLM fine-tuned on AR-specific data, or video question answering for dynamic content, could move part of the semantic evaluation from the cloud to the edge and shrink the 9.62 s manipulation latency; the paper notes this direction but does not implement it.
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

4 major / 5 minor

Summary. The paper formalizes two types of task-detrimental AR content—obstruction attacks and information manipulation attacks—and proposes ViDDAR, a user-edge-cloud system that uses a cloud VLM (GPT-4o or LLaVA-Next-8b) to identify key objects, an edge object detector and segmenter (Grounding DINO and SAM) to compute masks, and a pixel-level comparison to detect obstruction. For information manipulation, the VLM evaluates alignment, style similarity, and information misrepresentation. The system is evaluated on self-collected datasets (306 and 114 image pairs, respectively) and on a real-time Android AR app, reporting up to 92.15% obstruction detection accuracy at 533 ms latency and 82.46% information manipulation accuracy at 9.62 s. The authors also release the dataset and report an IRB-approved user study validating the labels.

Significance. If the results hold, ViDDAR would be the first VLM-based detector for task-detrimental AR content, and the formal framework, open dataset, and user study are valuable contributions to the AR safety community. The comparison against traditional baselines (saliency, Canny edge) and the ablation-style baselines (end-to-end, underdetailed, greedy) provides useful evidence for the design choices. However, the evaluation's single-object test regime, post-hoc threshold selection, and lack of statistical rigor mean that the general claims of robust task-detrimental-content detection are not yet fully established.

major comments (4)
  1. [Section 5.1.1, Eq. (4)] The formal obstruction definition triggers when any key object in K is obstructed, but the implemented pipeline asks the VLM to 'Give only one object' and the obstruction dataset contains exactly one key object per image pair. As a result, the reported 92.15% accuracy never exercises the existential quantifier, and in a scene with two task-critical objects the system would miss an obstruction of the object not selected by the VLM. The Greedy baseline is the only variant that enumerates all candidates, yet it is not ViDDAR and its accuracy is lower (88.89%); the paper therefore does not support the general claim of task-detrimental-content detection for multi-object scenes.
  2. [Section 5.1.1, Fig. 8(a)] The obstruction threshold α = 0.25 is chosen post-hoc from the dataset's 'approximate boundary' between obstructed and non-obstructed images, and the same dataset is then used to report the 92.15% accuracy. This makes the reported accuracy partly self-fulfilling; the paper should provide a sensitivity analysis over α, a held-out validation set, or independent labeling of the boundary to show that the threshold is not overfit to the evaluation set.
  3. [Section 5.1.2, Table 1, and Section 6] The paper acknowledges VLM response randomness, yet all accuracy numbers are point estimates from a single run with no confidence intervals or repeated trials. The gap between ViDDAR (92.15%) and the prior-knowledge upper bound (93.14%) is smaller than the likely run-to-run variability, so the claim that ViDDAR approaches the upper bound is not statistically supported. Reporting means and standard deviations over multiple runs (or otherwise quantifying randomness) is necessary to support the headline accuracies.
  4. [Section 5.2.1, Eq. (7)] The information manipulation ground truth is defined as M = A ∧ S ∧ I, and the VLM prompt (Question 6) instructs the model to answer 'yes' exactly when all three subquestions are 'yes'. The reported 82.46% accuracy therefore measures the VLM's internal consistency with the authors' rule rather than its ability to detect manipulation against an independent standard. Although the user study validates the factor labels, it does not provide an independent gold standard for the conjunction M; an external benchmark or human judgments on the test images would be needed to establish that the VLM is detecting manipulation rather than simply reproducing the labeling rule.
minor comments (5)
  1. [Section 4.1] The key-object list initialization is described as 'initiated manually by the user' with no automatic refresh; this operational limitation should be discussed in the limitations section, as it affects the claimed real-time applicability of the system.
  2. [Table 3] Latency results are reported as mean values only; please include standard deviations and clarify the 'six-hop' network configuration, which is not defined in the text.
  3. [Section 5.3] The user study uses only 10 image pairs per task and 20 participants; reporting per-item Likert scores and inter-rater agreement (e.g., Cohen's kappa) would strengthen the claim that the labels align with human perception.
  4. [References] The text cites 'GPT-4o-2024-08-06' but reference [31] is the GPT-4 technical report; please clarify the exact model version and access date.
  5. [Throughout] There are several typos and inconsistencies, e.g., 'VIDDAR' in Section 4.1 headers and 'ViDDAR' elsewhere; a careful proofread is needed.

Circularity Check

1 steps flagged · score 5.0 of 10

Obstruction threshold α is fitted to the dataset's label boundary before reporting 92.15% accuracy; the remainder of the evaluation is largely independent.

  1. fitted input called prediction [Section 5.1.1, Experiment Setup (after Eq. 4 and dataset description)]
    "During analysis, we identified that the approximate boundary between "obstructed" and "not obstructed" images was 0.25, so we set the obstruction threshold α = 0.25."

    The obstruction detector's decision rule (Eq. 4) is a thresholded overlap ratio: O = 1 iff |m_i^k ∩ m_c| ≥ α |m_i^k|. The threshold α is a free parameter of the detector. The paper sets α to the empirically observed boundary between the manually labeled 'obstructed' and 'not obstructed' images in the same dataset that is later used to report 92.15% obstruction detection accuracy. Thus, the detector's classification threshold is fitted to the ground-truth labels of the evaluation set, and the reported accuracy is computed on that same set. This is a form of fitting the decision rule to the test labels, so the 92.15% figure is partly self-fulfilling and does not provide an unbiased estimate of generalization to new scenes.

full rationale

ViDDAR is a detection system with a proposed formal model, and most of its evaluation is independent of the model's definitions. The information manipulation definition M = A ∧ S ∧ I is used both in the dataset labeling and in the VLM prompt's final question; however, this is an implementation of the paper's own definition rather than a circular derivation, and the ground-truth labels are independently validated by a 20-participant IRB-approved user study. The obstruction detection pipeline's key object recognition, segmentation, and overlap computation are all measured against human-validated labels, and the comparison to prior-knowledge, end-to-end, underdetailed, greedy, saliency, and Canny baselines provides independent evidence of relative performance. The one clear circular step is the choice of obstruction threshold α = 0.25: it is identified from the boundary between manually labeled obstructed and non-obstructed images in the same dataset used for the reported 92.15% accuracy. This fits the detector's decision threshold to the evaluation labels, so the reported accuracy is partly self-fulfilling and not an unbiased estimate of generalization. No load-bearing self-citations or imported uniqueness theorems are present.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

The central contribution is a pipeline built from existing models; the main fitted parameter is alpha. The mathematical framework is definitional, and the information manipulation model is an ad hoc AND-combination of three subjective factors. No new physical or conceptual entities are introduced beyond the formalized attack definitions.

free parameters (1)
  • alpha = 0.25
    Obstruction threshold in Eq. (4); chosen post-hoc in Section 5.1.1 based on the boundary observed in the collected dataset, so the reported accuracy depends on this fitted value.
assumptions (5)
  • domain assumption Key objects K are scenario-dependent and identifiable by a VLM from the raw image alone.
    Section 3.1 states K is not predefined and must be inferred; Section 4.1 uses a VLM prompt to identify exactly one key object. If the VLM fails, the whole obstruction pipeline fails.
  • ad hoc to paper Information manipulation occurs iff all three Boolean factors A, S, and I are true (Eq. 7).
    Section 3.2 introduces this as a definition without empirical justification; the user study only validates that labels agree with human perception, not that the AND combination is the necessary and sufficient condition.
  • domain assumption Grounding DINO and SAM generalize to AR scenes with virtual overlays.
    Section 4.1 uses these models without fine-tuning; their performance on AR images is not independently evaluated, and the reported segmentation mIoU (72%) indicates imperfect object masks.
  • domain assumption The full-reference setup assumes access to both raw and augmented images (Ir, Ia).
    Section 3.1 and the architecture require both images; many closed-source AR apps do not expose Ia, limiting deployment outside the authors' own app.
  • domain assumption The user study Likert responses validate the dataset labeling.
    Section 5.3 assumes that 20 participants' agreement (average >4) is sufficient evidence that labels align with human perception, but the study uses leading statements and a small sample.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ViDDAR: Vision Language Model-Based Task-Detrimental Content Detection for Augmented Reality." pith.science (2026). https://pith.science/paper/OU334ZYP

@misc{pith2026250112553,
  author       = {Pith},
  title        = {Pith review of: ViDDAR: Vision Language Model-Based Task-Detrimental Content Detection for Augmented Reality},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OU334ZYP}},
  note         = {Machine review of arXiv:2501.12553}
}
read the original abstract

In Augmented Reality (AR), virtual content enhances user experience by providing additional information. However, improperly positioned or designed virtual content can be detrimental to task performance, as it can impair users' ability to accurately interpret real-world information. In this paper we examine two types of task-detrimental virtual content: obstruction attacks, in which virtual content prevents users from seeing real-world objects, and information manipulation attacks, in which virtual content interferes with users' ability to accurately interpret real-world information. We provide a mathematical framework to characterize these attacks and create a custom open-source dataset for attack evaluation. To address these attacks, we introduce ViDDAR (Vision language model-based Task-Detrimental content Detector for Augmented Reality), a comprehensive full-reference system that leverages Vision Language Models (VLMs) and advanced deep learning techniques to monitor and evaluate virtual content in AR environments, employing a user-edge-cloud architecture to balance performance with low latency. To the best of our knowledge, ViDDAR is the first system to employ VLMs for detecting task-detrimental content in AR settings. Our evaluation results demonstrate that ViDDAR effectively understands complex scenes and detects task-detrimental content, achieving up to 92.15% obstruction detection accuracy with a detection latency of 533 ms, and an 82.46% information manipulation content detection accuracy with a latency of 9.62 s.

Figures

Figures reproduced from arXiv: 2501.12553 by the authors.

Figure 1
Figure 1. Example of an obstruction attack in AR: (a) real-world view; (b) [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Example of an information manipulation attack in AR: (a) real [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Whether an object is considered a key object depends on the [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (6 more)
Figure 5
Figure 5. Figure 5: System architecture of ViDDAR for obstruction detection. [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]
Figure 6
Figure 6. Figure 6: System architecture of ViDDAR for information manipulation [PITH_FULL_IMAGE:figures/full_fig_p005_6.png]
Figure 7
Figure 7. Figure 7: Obstruction attack dataset samples. The first row shows raw images; the second row shows the augmented images; the third row shows the [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]
Figure 8
Figure 8. Figure 8: Statistical analysis of datasets we collected: (a): Distribution of [PITH_FULL_IMAGE:figures/full_fig_p006_8.png]
Figure 9
Figure 9. Figure 9: Information manipulation attack dataset samples. The first row shows raw images; the second row shows the augmented images. Their labels [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]
Figure 10
Figure 10. Figure 10: An example of statements used in the information manipulation [PITH_FULL_IMAGE:figures/full_fig_p009_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 41 canonical work pages

  1. [1]

    Introducing the next generation of Claude

    Anthropic. Introducing the next generation of Claude. https://www. anthropic.com/news/claude-3-family , 2024. 3

    work page 2024

  2. [2]

    Bosse, D

    S. Bosse, D. Maniry, K. Müller, T. Wiegand, and W. Samek. Deep neural networks for no-reference and full-reference image quality assessment. IEEE Transactions on Image Processing, 27(1):206–219, 2018. 1

    work page 2018

  3. [3]

    J. Canny. A computational approach to edge detection. IEEE Transactions on Pattern Analysis and Machine Intelligence, (6):679–698, 1986. 5

    work page 1986

  4. [4]

    D. M. Chandler. Seven challenges in image quality assessment: past, present, and future research. International Scholarly Research Notices, 2013(1):905685, 2013. 1

    work page 2013

  5. [5]

    Cheng, J

    K. Cheng, J. F. Tian, T. Kohno, and F. Roesner. Exploring user reactions and mental models towards perceptual manipulation attacks in mixed reality. In Proceedings of USENIX Security, 2023. 1

    work page 2023

  6. [6]

    Davari, F

    S. Davari, F. Lu, and D. A. Bowman. Occlusion management techniques for everyday glanceable AR interfaces. In IEEE VR Abstracts and Work- shops (VRW), 2020. 1, 2

    work page 2020

  7. [7]

    K. Ding, K. Ma, S. Wang, and E. P. Simoncelli. Comparison of full- reference image quality models for optimization of image processing systems. International Journal of Computer Vision, 129(4):1258–1281,

    work page

  8. [8]

    K. Ding, K. Ma, S. Wang, and E. P. Simoncelli. Image quality assessment: Unifying structure and texture similarity. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(5):2567–2581, 2022. 1

    work page 2022

Show all 47 references
  1. [9]

    Dosovitskiy, L

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Un- terthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale. In Proceedings of ICLR, 2021. 3

    2021

  2. [10]

    H. Duan, L. Guo, W. Sun, X. Min, L. Chen, and G. Zhai. Augmented reality image quality assessment based on visual confusion theory. In2022 IEEE International Symposium on BMSB, pp. 1–6, 2022. 2

    2022

  3. [11]

    H. Duan, X. Min, Y . Zhu, G. Zhai, X. Yang, and P. Le Callet. Confusing image quality assessment: Toward better augmented reality experience. IEEE Transactions on Image Processing, 31:7206–7221, 2022. 1, 2

    2022

  4. [12]

    H. Duan, W. Shen, X. Min, D. Tu, J. Li, and G. Zhai. Saliency in aug- mented reality. In Proceedings of the ACM International Conference on Multimedia, p. 6549–6558, 2022. 2

    2022

  5. [13]

    Eghtebas, G

    C. Eghtebas, G. Klinker, S. Boll, and M. Koelle. Co-speculating on dark scenarios and unintended consequences of a ubiquitous(ly) augmented reality. In Proceedings of the ACM Designing Interactive Systems Confer- ence, p. 2392–2407, 2023. 2

    2023

  6. [14]

    J. L. Gabbard, J. E. Swan, D. Hix, S.-J. Kim, and G. Fitch. Active text drawing styles for outdoor augmented reality: A user-based study and design implications. In Proceedings of IEEE VR, pp. 35–42, 2007. 3

    2007

  7. [15]

    Gemini: A family of highly capable multimodal models

    Gemini Team and ik. Gemini: A family of highly capable multimodal models. arXiv:2312.11805, 2024. 3

    2024 arXiv

  8. [16]

    T. Guan, F. Liu, X. Wu, R. Xian, Z. Li, X. Liu, X. Wang, L. Chen, F. Huang, Y . Yacoob, D. Manocha, and T. Zhou. HallusionBench: An advanced diagnostic suite for entangled language hallucination and visual illusion in large vision-language models. In Proceedings of the IEEE/CV...

    2024

  9. [17]

    Hou and L

    X. Hou and L. Zhang. Saliency detection: A spectral residual approach. In Proceedings of IEEE/CVF CVPR, pp. 1–8, 2007. 1, 2, 5

    2007

  10. [18]

    Kirillov, E

    A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y . Lo, P. Dollár, and R. Girshick. Segment anything. arXiv:2304.02643, 2023. 4

    2023 arXiv

  11. [19]

    A. Kong, S. Zhao, H. Chen, Q. Li, Y . Qin, R. Sun, X. Zhou, E. Wang, and X. Dong. Better zero-shot reasoning with role-play prompting. arXiv:2308.07702, 2024. 5

    2024 arXiv

  12. [20]

    Langlotz, T

    T. Langlotz, T. Nguyen, D. Schmalstieg, and R. Grasset. Next-generation augmented reality browsers: Rich, seamless, and adaptive. Proceedings of the IEEE, 102(2):155–169, 2014. 3

    2014

  13. [21]

    Lebeck, K

    K. Lebeck, K. Ruth, T. Kohno, and F. Roesner. Securing augmented reality output. In Proceedings of IEEE Symposium on Security and Privacy (SP), pp. 320–337, 2017. 1, 2

    2017

  14. [22]

    Lebeck, K

    K. Lebeck, K. Ruth, T. Kohno, and F. Roesner. Arya: Operating sys- tem support for securely augmenting reality. IEEE Security & Privacy, 16(01):44–53, 2018. 1, 2

    2018

  15. [23]

    U. Lee, H. Jung, Y . Jeon, Y . Sohn, W. Hwang, J. Moon, and H. Kim. Few-shot is enough: exploring ChatGPT prompt engineering method for automatic question generation in English education. Education and Information Technologies, 29(9):11483–11515, 2024. 5

    2024

  16. [24]

    R. Likert. A technique for the measurement of attitudes. Archives of Psychology, 1932. 8

    1932

  17. [25]

    H. Liu, C. Li, Q. Wu, and Y . J. Lee. Visual instruction tuning. InProceed- ings of NeurIPS, vol. 36, pp. 34892–34916, 2023. 3, 4

    2023

  18. [26]

    S. Liu, Z. Zeng, T. Ren, F. Li, H. Zhang, J. Yang, C. Li, J. Yang, H. Su, J. Zhu, et al. Grounding DINO: Marrying DINO with grounded pre- training for open-set object detection. arXiv:2303.05499, 2023. 4

    2023 arXiv

  19. [27]

    Z. Liu, L. Zhu, B. Shi, Z. Zhang, Y . Lou, S. Yang, H. Xi, S. Cao, Y . Gu, D. Li, X. Li, Y . Fang, Y . Chen, C.-Y . Hsieh, D.-A. Huang, A.-C. Cheng, V . Nath, J. Hu, S. Liu, R. Krishna, D. Xu, X. Wang, P. Molchanov, J. Kautz, H. Yin, S. Han, and Y . Lu. NVILA: Efficient fronti...

    2024 arXiv

  20. [28]

    J. Lu, D. Batra, D. Parikh, and S. Lee. ViLBERT: Pretraining task- agnostic visiolinguistic representations for vision-and-language tasks. In Proceedings of NeurIPS, vol. 32, 2019. 1

    2019

  21. [29]

    W. Luo, A. Lehmann, H. Widengren, and R. Dachselt. Where should we put it? Layout and placement strategies of documents in augmented reality for collaborative sensemaking. In Proceedings of the CHI Conference on Human Factors in Computing Systems, 2022. 3

    2022

  22. [30]

    Nuernberger, E

    B. Nuernberger, E. Ofek, H. Benko, and A. D. Wilson. SnapToReality: Aligning augmented reality to the real world. In Proceedings of CHI Conference on Human Factors in Computing Systems , p. 1233–1244,

  23. [31]

    GPT-4 technical report

    OpenAI. GPT-4 technical report. arXiv:2303.08774, 2023. 3, 4, 5

    2023 arXiv

  24. [32]

    Radford, J

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark, G. Krueger, and I. Sutskever. Learning transferable visual models from natural language supervision. In Proceedings of the International Conference on Machine Learning...

    2021

  25. [33]

    E. E. Sabelman and R. Lam. The real-life dangers of augmented reality. IEEE Spectrum, 52(7):48–53, 2015. 1

    2015

  26. [34]

    Satkowski, R

    M. Satkowski, R. Rzayev, E. Goebel, and R. Dachselt. ABOVE & BELOW: Investigating ceiling and floor for augmented reality content placement. In Proceedings of IEEE ISMAR, 2022. 1, 2

    2022

  27. [35]

    M. M. Shah, H. Arshad, and R. Sulaiman. Occlusion in augmented reality. In Proceedings of International Conference on Information Science and Digital Content Technology, vol. 2, pp. 372–378, 2012. 1, 2

    2012

  28. [36]

    Sheikh and A

    H. Sheikh and A. Bovik. Image information and visual quality. IEEE Transactions on Image Processing, 15(2):430–444, 2006. 1

    2006

  29. [37]

    Y . Shen, J. Shao, X. Zhang, Z. Lin, H. Pan, D. Li, J. Zhang, and K. B. Letaief. Large language models empowered autonomous edge AI for connected intelligence. IEEE Communications Magazine, 2024. 9

    2024

  30. [38]

    S. Tong, Z. Liu, Y . Zhai, Y . Ma, Y . LeCun, and S. Xie. Eyes wide shut? Exploring the visual shortcomings of multimodal LLMs. In Proceedings of the IEEE/CVF CVPR, pp. 9568–9578, June 2024. 6

    2024

  31. [39]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin. Attention is all you need. In Proceedings of NeurIPS, p. 6000–6010, 2017. 3

    2017

  32. [40]

    Wang, L.-H

    X. Wang, L.-H. Lee, C. Bermejo Fernandez, and P. Hui. The dark side of augmented reality: Exploring manipulative designs in AR. International Journal of Human–Computer Interaction, 40(13):3449–3464, 2024. 2

    2024

  33. [41]

    Z. Wang, A. Bovik, H. Sheikh, and E. Simoncelli. Image quality assess- ment: from error visibility to structural similarity. IEEE Transactions on Image Processing, 13(4):600–612, 2004. 1

    2004

  34. [42]

    Z. Wang, E. Simoncelli, and A. Bovik. Multiscale structural similarity for image quality assessment. In Proceedings of Asilomar Conference on Signals, Systems and Computers, vol. 2, pp. 1398–1402 V ol.2, 2003. 1

    2003

  35. [43]

    Y . Xiu, T. Scargill, and M. Gorlatova. LOBSTAR: Language model-based obstruction detection for augmented reality. In Proceedings of IEEE ISMAR-Adjunct, 2024. 1

    2024

  36. [44]

    S. Yu, J. Cho, P. Yadav, and M. Bansal. Self-chained image-language model for video localization and question answering. In A. Oh, T. Nau- mann, A. Globerson, K. Saenko, M. Hardt, and S. Levine, eds., Proceed- ings of NeurIPS, vol. 36, pp. 76749–76771, 2023. 9

    2023

  37. [45]

    Zhang, S

    F. Zhang, S. Li, L. Ma, and K. N. Ngan. Limitation and challenges of image quality measurement. In Proceedings of Visual Communications and Image Processing, vol. 7744, pp. 25–32. SPIE, 2010. 1

    2010

  38. [46]

    Zhang, J

    J. Zhang, J. Huang, S. Jin, and S. Lu. Vision-language models for vision tasks: A survey. IEEE Transactions on Pattern Analysis and Machine Intelligence, 46(8):5625–5644, 2024. 1

    2024

  39. [47]

    Zhang, L

    L. Zhang, L. Zhang, X. Mou, and D. Zhang. A comprehensive evalu- ation of full reference image quality assessment algorithms. In IEEE International Conference on Image Processing, pp. 1477–1480, 2012. 3

    2012

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.