pith. sign in

arxiv: 2606.09132 · v1 · pith:Y6U7RAR4new · submitted 2026-06-08 · 💻 cs.AI

Vision Language Model Helps Private Information De-Identification in Vision Data

Tiejin Chen , Pingzhi Li , Kaixiong Zhou , Tianlong Chen , Hua Wei This is my paper

Pith reviewed 2026-06-27 16:27 UTC · model grok-4.3

classification 💻 cs.AI
keywords vision-language modelsprivacy de-identificationsensitive text localizationOPTIC datasetbounding boxesprotected health informationoptical character recognitionimage privacy
0
0 comments X

The pith

VisShield trains vision-language models to localize and mask private text in images using a specialized dataset.

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

The paper introduces VisShield, an end-to-end framework that adapts vision-language models to privacy tasks in visual data. It creates the OPTIC dataset with privacy-oriented prompts to guide targeted optical character recognition and uses a training method so that models output bounding boxes around sensitive text for subsequent masking. This targets risks like protected health information in medical images, which existing privacy methods for text have overlooked in visual inputs. The authors report that experiments show the approach outperforms prior methods at handling private information. If correct, it supplies a concrete way for popular multimodal models to support de-identification workflows.

Core claim

The central claim is that the VisShield framework, built from the OPTIC instruction-tuning dataset and a tailored training methodology, lets vision-language models recognize privacy-sensitive text, perform precise localization, and output bounding boxes that enable effective masking, thereby outperforming existing approaches in privacy protection for vision data.

What carries the argument

The VisShield framework consisting of the OPTIC dataset that supplies privacy-oriented prompts for targeted OCR and the training strategy that adapts VLMs to output bounding boxes for sensitive entities.

Load-bearing premise

The assumption that instruction tuning on the OPTIC dataset and the tailored training methodology will enable VLMs to accurately localize sensitive text and produce usable bounding boxes for effective masking.

What would settle it

An experiment in which the adapted VLM produces bounding boxes that fail to align with ground-truth sensitive text locations or shows no improvement over baselines on privacy metrics would falsify the central claim.

Figures

Figures reproduced from arXiv: 2606.09132 by Hua Wei, Kaixiong Zhou, Pingzhi Li, Tianlong Chen, Tiejin Chen.

Figure 1
Figure 1. Figure 1: An illustrative example of medical imaging [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The proposed de-identification pipeline. Our approach leverages instruction-tuned VLMs to first perform [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Overview of our three-stage dataset gener [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Template prompt utilized for instruction generation, implemented with GPT-4 and Claude-3.5 Sonnet. [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: IoU performance comparison with different [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: An example of de-identification of private [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: One instruction prompt example generated by GPT-4o. [PITH_FULL_IMAGE:figures/full_fig_p014_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: IoU performance comparison with different [PITH_FULL_IMAGE:figures/full_fig_p015_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: IoU performance comparison with different [PITH_FULL_IMAGE:figures/full_fig_p015_9.png] view at source ↗
read the original abstract

Visual Language Models (VLMs) have gained significant popularity due to their remarkable ability. While various methods exist to enhance privacy in text-based applications, privacy risks associated with visual inputs remain largely overlooked such as Protected Health Information (PHI) in medical images. To tackle this problem, two key tasks: accurately localizing sensitive text and processing it to ensure privacy protection should be performed. To address this issue, we introduce VisShield (Vision Privacy Shield), an end-to-end framework designed to enhance the privacy awareness of VLMs. Our framework consists of two key components: a specialized instruction-tuning dataset OPTIC (Optical Privacy Text Instruction Collection) and a tailored training methodology. The dataset provides diverse privacy-oriented prompts that guide VLMs to perform targeted Optical Character Recognition (OCR) for precise localization of sensitive text, while the training strategy ensures effective adaptation of VLMs to privacy-preserving tasks. Specifically, our approach ensures that VLMs recognize privacy-sensitive text and output precise bounding boxes for detected entities, allowing for effective masking of sensitive information. Extensive experiments demonstrate that our framework significantly outperforms existing approaches in handling private information, paving the way for privacy-preserving applications in vision-language models. Our dataset and code can be found here.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces VisShield, an end-to-end framework that instruction-tunes vision-language models on the OPTIC dataset to perform targeted OCR, output bounding boxes around sensitive text (e.g., PHI), and enable subsequent masking. It claims this yields significant outperformance over existing approaches for privacy preservation in vision data.

Significance. If the localization step is shown to be reliable, the framework could meaningfully advance privacy-preserving VLM applications in domains such as medical imaging. Releasing the OPTIC dataset and code is a positive contribution to reproducibility.

major comments (2)
  1. [Section 4] Section 4: Experiments report only downstream privacy metrics (de-identification success rates) and qualitative examples. No tables or text supply standard localization metrics such as mean IoU, precision@IoU=0.5, or recall for the bounding-box outputs on held-out images. This is load-bearing because the central claim requires that the boxes be precise enough for effective masking; without these numbers it is impossible to distinguish reliable localization from prompt-engineering effects.
  2. [Section 3] Section 3: The training objective is described as eliciting OCR plus box output, yet no ablation or validation is provided on how well the fine-tuned model generalizes to unseen image distributions or prompt variations that would affect box quality.
minor comments (2)
  1. [Abstract] Abstract and Section 1 should explicitly list the baselines, metrics, and error analysis used in the experiments rather than stating only that outperformance was observed.
  2. [Section 3] Notation for bounding-box coordinates and the masking procedure should be defined once with consistent symbols across Sections 3 and 4.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thoughtful review and constructive suggestions. We address the major comments below and plan to incorporate revisions to strengthen the manuscript.

read point-by-point responses

  1. Referee: Section 4: Experiments report only downstream privacy metrics (de-identification success rates) and qualitative examples. No tables or text supply standard localization metrics such as mean IoU, precision@IoU=0.5, or recall for the bounding-box outputs on held-out images. This is load-bearing because the central claim requires that the boxes be precise enough for effective masking; without these numbers it is impossible to distinguish reliable localization from prompt-engineering effects.

    Authors: We agree that providing standard localization metrics would offer a more comprehensive evaluation of the bounding box outputs. Although the de-identification success rates indirectly validate the localization quality (as inaccurate boxes would lead to failed masking), we will revise Section 4 to include a table with mean IoU, precision at IoU=0.5, and recall computed on held-out images. This will help demonstrate that the localization is reliable rather than due to prompt engineering. revision: yes

  2. Referee: Section 3: The training objective is described as eliciting OCR plus box output, yet no ablation or validation is provided on how well the fine-tuned model generalizes to unseen image distributions or prompt variations that would affect box quality.

    Authors: The comment is valid; additional validation on generalization would be beneficial. We will add experiments evaluating the model on unseen image distributions (e.g., non-medical images or different scanners) and different prompt variations, reporting the corresponding localization metrics to assess robustness. revision: yes

Circularity Check

0 steps flagged

No derivation chain or self-referential elements present

full rationale

The paper presents an applied framework (VisShield) built around a new instruction-tuning dataset (OPTIC) and a training procedure for VLMs to output bounding boxes for sensitive text, followed by masking. No equations, parameters fitted to subsets then re-predicted, or mathematical derivations appear in the abstract or described sections. Claims rest on downstream empirical results rather than any reduction to prior self-citations or definitional loops. The absence of any load-bearing derivation chain means the work is self-contained against external benchmarks with no circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only abstract available; no free parameters, axioms, or invented entities identifiable.

pith-pipeline@v0.9.1-grok · 5750 in / 950 out tokens · 27015 ms · 2026-06-27T16:27:24.696580+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

84 extracted references · 38 canonical work pages · 18 internal anchors

  1. [1]

    Advances in neural information processing systems , volume=

    Visual instruction tuning , author=. Advances in neural information processing systems , volume=

  2. [2]

    International conference on machine learning , pages=

    Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models , author=. International conference on machine learning , pages=. 2023 , organization=

    2023

  3. [3]

    Advances in neural information processing systems , volume=

    Flamingo: a visual language model for few-shot learning , author=. Advances in neural information processing systems , volume=

  4. [4]

    Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution

    Qwen2-vl: Enhancing vision-language model's perception of the world at any resolution , author=. arXiv preprint arXiv:2409.12191 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  5. [5]

    arXiv preprint arXiv:2405.02246 , year=

    What matters when building vision-language models? , author=. arXiv preprint arXiv:2405.02246 , year=

    work page arXiv

  6. [6]

    European Conference on Computer Vision , pages=

    Vary: Scaling up the vision vocabulary for large vision-language model , author=. European Conference on Computer Vision , pages=. 2025 , organization=

    2025

  7. [7]

    PaliGemma: A versatile 3B VLM for transfer

    Paligemma: A versatile 3b vlm for transfer , author=. arXiv preprint arXiv:2407.07726 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  8. [8]

    arXiv preprint arXiv:2405.20797 , year=

    Ovis: Structural Embedding Alignment for Multimodal Large Language Model , author=. arXiv preprint arXiv:2405.20797 , year=

    work page arXiv

  9. [9]

    CogVLM: Visual Expert for Pretrained Language Models

    Cogvlm: Visual expert for pretrained language models , author=. arXiv preprint arXiv:2311.03079 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  10. [10]

    MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models

    Minigpt-4: Enhancing vision-language understanding with advanced large language models , author=. arXiv preprint arXiv:2304.10592 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  11. [11]

    DeepSeek-VL: Towards Real-World Vision-Language Understanding

    Deepseek-vl: towards real-world vision-language understanding , author=. arXiv preprint arXiv:2403.05525 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  12. [12]

    arXiv preprint arXiv:2408.08872 , year=

    xgen-mm (blip-3): A family of open large multimodal models , author=. arXiv preprint arXiv:2408.08872 , year=

    work page arXiv

  13. [13]

    arXiv preprint arXiv:2309.11499 , year=

    Dreamllm: Synergistic multimodal comprehension and creation , author=. arXiv preprint arXiv:2309.11499 , year=

    work page arXiv

  14. [14]

    arXiv preprint arXiv:2307.09474 , year=

    Chatspot: Bootstrapping multimodal llms via precise referring instruction tuning , author=. arXiv preprint arXiv:2307.09474 , year=

    work page arXiv

  15. [15]

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

    Latr: Layout-aware transformer for scene-text vqa , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  16. [16]

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

    From images to textual prompts: Zero-shot visual question answering with frozen large language models , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  17. [17]

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

    Enhancing Visual Question Answering through Question-Driven Image Captions as Prompts , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  18. [18]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Bliva: A simple multimodal llm for better handling of text-rich visual questions , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  19. [19]

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

    Fusecap: Leveraging large language models for enriched fused image captions , author=. Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , pages=

  20. [20]

    Advances in Neural Information Processing Systems , volume=

    Exploring diverse in-context configurations for image captioning , author=. Advances in Neural Information Processing Systems , volume=

  21. [21]

    International Conference on Machine Learning , pages=

    Pix2struct: Screenshot parsing as pretraining for visual language understanding , author=. International Conference on Machine Learning , pages=. 2023 , organization=

    2023

  22. [22]

    Kosmos-2: Grounding Multimodal Large Language Models to the World

    Kosmos-2: Grounding multimodal large language models to the world , author=. arXiv preprint arXiv:2306.14824 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  23. [23]

    European Conference on Computer Vision , pages=

    Merlin: Empowering multimodal llms with foresight minds , author=. European Conference on Computer Vision , pages=. 2025 , organization=

    2025

  24. [24]

    arXiv preprint arXiv:2309.11419 , year=

    Kosmos-2.5: A multimodal literate model , author=. arXiv preprint arXiv:2309.11419 , year=

    work page arXiv

  25. [25]

    Advances in neural information processing systems , volume=

    Training language models to follow instructions with human feedback , author=. Advances in neural information processing systems , volume=

  26. [26]

    arXiv preprint arXiv:2204.07705 , volume=

    Benchmarking generalization via in-context instructions on 1,600+ language tasks , author=. arXiv preprint arXiv:2204.07705 , volume=

    work page arXiv

  27. [27]

    Finetuned Language Models Are Zero-Shot Learners

    Finetuned language models are zero-shot learners , author=. arXiv preprint arXiv:2109.01652 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  28. [28]

    arXiv preprint arXiv:2308.10792 , year=

    Instruction tuning for large language models: A survey , author=. arXiv preprint arXiv:2308.10792 , year=

    work page arXiv

  29. [29]

    GPT-4 Technical Report

    GPT-4 Technical Report , author=. arXiv preprint arXiv:2303.08774, 2023 , year=

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  30. [30]

    2023 , url =

    OpenAI , title =. 2023 , url =

    2023

  31. [31]

    Llama 2: Open Foundation and Fine-Tuned Chat Models

    Llama 2: Open foundation and fine-tuned chat models , author=. arXiv preprint arXiv:2307.09288 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  32. [32]

    The Llama 3 Herd of Models

    The llama 3 herd of models , author=. arXiv preprint arXiv:2407.21783 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  33. [33]

    International Conference on Machine Learning , pages=

    The flan collection: Designing data and methods for effective instruction tuning , author=. International Conference on Machine Learning , pages=. 2023 , organization=

    2023

  34. [34]

    Journal of Machine Learning Research , volume=

    Scaling instruction-finetuned language models , author=. Journal of Machine Learning Research , volume=

  35. [35]

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

    Improved baselines with visual instruction tuning , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  36. [36]

    Advances in Neural Information Processing Systems , volume=

    Lima: Less is more for alignment , author=. Advances in Neural Information Processing Systems , volume=

  37. [37]

    Computer Vision--ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13 , pages=

    Microsoft coco: Common objects in context , author=. Computer Vision--ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13 , pages=. 2014 , organization=

    2014

  38. [38]

    arXiv preprint arXiv:2306.17107 , year=

    Llavar: Enhanced visual instruction tuning for text-rich image understanding , author=. arXiv preprint arXiv:2306.17107 , year=

    work page arXiv

  39. [39]

    arXiv 2023 , author=

    Shikra: Unleashing Multimodal LLM’s Referential Dialogue Magic. arXiv 2023 , author=

    2023

  40. [40]

    InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning , url =

    Dai, Wenliang and Li, Junnan and LI, DONGXU and Tiong, Anthony and Zhao, Junqi and Wang, Weisheng and Li, Boyang and Fung, Pascale N and Hoi, Steven , booktitle =. InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning , url =

  41. [41]

    Proceedings of the IEEE international conference on computer vision , pages=

    Flickr30k entities: Collecting region-to-phrase correspondences for richer image-to-sentence models , author=. Proceedings of the IEEE international conference on computer vision , pages=

  42. [42]

    Scientific Data , volume=

    A DICOM dataset for evaluation of medical image de-identification , author=. Scientific Data , volume=. 2021 , publisher=

    2021

  43. [43]

    Signal Processing: Image Communication , volume=

    De-identification for privacy protection in multimedia content: A survey , author=. Signal Processing: Image Communication , volume=. 2016 , publisher=

    2016

  44. [44]

    Proceedings of the Workshop on NLP and Pseudonymisation , volume=

    Pseudonymisation of Swedish electronic patient records using a rule-based approach , author=. Proceedings of the Workshop on NLP and Pseudonymisation , volume=

  45. [45]

    Journal of Personalized medicine , volume=

    Verification of de-identification techniques for personal information using tree-based methods with Shapley values , author=. Journal of Personalized medicine , volume=. 2022 , publisher=

    2022

  46. [46]

    Biomedical engineering systems and technologies, international joint conference, BIOSTEC

    DeIDNER Model: A Neural Network Named Entity Recognition Model for Use in the De-identification of Clinical Notes , author=. Biomedical engineering systems and technologies, international joint conference, BIOSTEC... revised selected papers. BIOSTEC (Conference) , volume=. 2022 , organization=

    2022

  47. [47]

    arXiv preprint arXiv:2303.11032 , year=

    Deid-gpt: Zero-shot medical text de-identification by gpt-4 , author=. arXiv preprint arXiv:2303.11032 , year=

    work page arXiv

  48. [48]

    2006 Conference on computer vision and pattern recognition workshop (CVPRW'06) , pages=

    Model-based face de-identification , author=. 2006 Conference on computer vision and pattern recognition workshop (CVPRW'06) , pages=. 2006 , organization=

    2006

  49. [49]

    2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) , pages=

    I know that person: Generative full body and face de-identification of people in images , author=. 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) , pages=. 2017 , organization=

    2017

  50. [50]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Personalized and invertible face de-identification by disentangled identity information manipulation , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  51. [51]

    Advances in Neural Information Processing Systems , volume=

    Revisiting resnets: Improved training and scaling strategies , author=. Advances in Neural Information Processing Systems , volume=

  52. [52]

    IEEE intelligent systems , volume=

    The unreasonable effectiveness of data , author=. IEEE intelligent systems , volume=. 2009 , publisher=

    2009

  53. [53]

    2009 IEEE conference on computer vision and pattern recognition , pages=

    Imagenet: A large-scale hierarchical image database , author=. 2009 IEEE conference on computer vision and pattern recognition , pages=. 2009 , organization=

    2009

  54. [54]

    Scaling Laws for Neural Language Models

    Scaling laws for neural language models , author=. arXiv preprint arXiv:2001.08361 , year=

    work page internal anchor Pith review Pith/arXiv arXiv 2001

  55. [55]

    OPT: Open Pre-trained Transformer Language Models

    Opt: Open pre-trained transformer language models , author=. arXiv preprint arXiv:2205.01068 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  56. [56]

    Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond

    Qwen-vl: A frontier large vision-language model with versatile abilities , author=. arXiv preprint arXiv:2308.12966 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  57. [57]

    arXiv preprint arXiv:2402.02103 , year=

    D 'ej a Vu Memorization in Vision-Language Models , author=. arXiv preprint arXiv:2402.02103 , year=

    work page arXiv

  58. [58]

    International Conference on Medical Image Computing and Computer-Assisted Intervention , pages=

    Can llms’ tuning methods work in medical multimodal domain? , author=. International Conference on Medical Image Computing and Computer-Assisted Intervention , pages=. 2024 , organization=

    2024

  59. [59]

    2024 IEEE/ACM Conference on Connected Health: Applications, Systems and Engineering Technologies (CHASE) , pages=

    On large visual language models for medical imaging analysis: An empirical study , author=. 2024 IEEE/ACM Conference on Connected Health: Applications, Systems and Engineering Technologies (CHASE) , pages=. 2024 , organization=

    2024

  60. [60]

    arXiv preprint arXiv:2409.15256 , year=

    Behavioral Bias of Vision-Language Models: A Behavioral Finance View , author=. arXiv preprint arXiv:2409.15256 , year=

    work page arXiv

  61. [61]

    Journal of the American Medical Informatics Association , volume=

    BoB, a best-of-breed automated text de-identification system for VHA clinical documents , author=. Journal of the American Medical Informatics Association , volume=. 2013 , publisher=

    2013

  62. [62]

    A Survey on In-context Learning

    A survey on in-context learning , author=. arXiv preprint arXiv:2301.00234 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  63. [63]

    Advances in Neural Information Processing Systems , volume=

    What makes good examples for visual in-context learning? , author=. Advances in Neural Information Processing Systems , volume=

  64. [64]

    2024 , url =

    Joke, Edén and contributors , title =. 2024 , url =

    2024

  65. [65]

    2024 , url =

    Pillow , author =. 2024 , url =

    2024

  66. [66]

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

    Connecting pixels to privacy and utility: Automatic redaction of private information in images , author=. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages=

  67. [67]

    LoRA: Low-Rank Adaptation of Large Language Models

    Lora: Low-rank adaptation of large language models , author=. arXiv preprint arXiv:2106.09685 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  68. [68]

    arXiv preprint arXiv:2210.07903 , year=

    Text detection forgot about document OCR , author=. arXiv preprint arXiv:2210.07903 , year=

    work page arXiv

  69. [69]

    IEEE transactions on pattern analysis and machine intelligence , volume=

    Faster R-CNN: Towards real-time object detection with region proposal networks , author=. IEEE transactions on pattern analysis and machine intelligence , volume=. 2016 , publisher=

    2016

  70. [70]

    2023 , howpublished =

    Presidio - Open Source Data Protection and Privacy Engineering Platform , author =. 2023 , howpublished =

    2023

  71. [71]

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

    Scene parsing through ade20k dataset , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  72. [72]

    Medical Image Computing and Computer-Assisted Intervention--MICCAI 2013: 16th International Conference, Nagoya, Japan, September 22-26, 2013, Proceedings, Part II 16 , pages=

    Automated separation of binary overlapping trees in low-contrast color retinal images , author=. Medical Image Computing and Computer-Assisted Intervention--MICCAI 2013: 16th International Conference, Nagoya, Japan, September 22-26, 2013, Proceedings, Part II 16 , pages=. 2013 , organization=

    2013

  73. [73]

    arXiv preprint arXiv:2304.08109 , year=

    A comparative study between full-parameter and lora-based fine-tuning on chinese instruction data for instruction following large language model , author=. arXiv preprint arXiv:2304.08109 , year=

    work page arXiv

  74. [74]

    Decoupled Weight Decay Regularization

    Decoupled weight decay regularization , author=. arXiv preprint arXiv:1711.05101 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  75. [75]

    Gemini: A Family of Highly Capable Multimodal Models

    Gemini: a family of highly capable multimodal models , author=. arXiv preprint arXiv:2312.11805 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  76. [76]

    Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone

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

    work page internal anchor Pith review Pith/arXiv arXiv

  77. [77]

    30th USENIX Security Symposium (USENIX Security 21) , pages=

    Extracting training data from large language models , author=. 30th USENIX Security Symposium (USENIX Security 21) , pages=

  78. [78]

    arXiv preprint arXiv:2205.12506 , year=

    Memorization in nlp fine-tuning methods , author=. arXiv preprint arXiv:2205.12506 , year=

    work page arXiv

  79. [79]

    arXiv preprint arXiv:2205.12628 , year=

    Are Large Pre-Trained Language Models Leaking Your Personal Information? , author=. arXiv preprint arXiv:2205.12628 , year=

    work page arXiv

  80. [80]

    arXiv preprint arXiv:2410.22108 , year=

    Protecting privacy in multimodal large language models with mllmu-bench , author=. arXiv preprint arXiv:2410.22108 , year=

    work page arXiv

Showing first 80 references.