Large language models (LLMs) are increasingly integrated into daily life through conversational interfaces, processing user data via natural language inputs and exhibiting advanced reasoning capabilities, which raises new concerns about user control over privacy. While much research has focused on potential privacy risks, less attention has been paid to the data control mechanisms these platforms provide. This study examines six conversational LLM platforms, analyzing how they define and implement features for users to access, edit, delete, and share data. Our analysis reveals an emerging paradigm of data control in conversational LLM platforms, where user data is generated and derived through interaction itself, natural language enables flexible yet often ambiguous control, and multi-user interactions with shared data raise questions of co-ownership and governance. Based on these findings, we offer practical insights for platform developers, policymakers, and researchers to design more effective and usable privacy controls in LLM-powered conversational interactions.

Test-time few-shot object detection (FSOD) represents an innovative approach for identifying novel categories using a limited number of support examples, obviating the need for model fine-tuning. Despite advancements, existing FSOD methods, including our prior work, continue to grapple with challenges posed by domain/category shift and limited data availability. Building upon our previous research on test-time FSOD, this article proposes a novel dynamic prototype fusion network (PFN) to overcome these limitations. To mitigate the impact of the distribution shift, a dynamic prototype refinement method is introduced that updates prototypes from supporting images in an adaptive manner. Further, limited samples are mitigated through exhaustive exploitation of information within support images. Specifically, we design a dual-level multiscale information integration approach that effectively fuses information across different network layers and image scales, enhancing the model's discriminating capabilities. Additionally, a mask-based preprocessing technique harnesses segmentation labels on support samples, effectively suppressing the adverse impact of background noise on model accuracy. Notably, to align with the constraints of test-time scenarios, model parameters remain fixed during the configuration step, with only prototypes being updated each time users input novel supporting samples. As a result, our method achieves superior performance over existing state-of-the-art FSOD methods on multiple benchmarks, demonstrating remarkable potential in the realm of FSOD. The code is available at https://github.com/CatfishW/TIDEV2.

Personally identifiable information (PII) is a fundamental concept in privacy research and regulations. Understanding users' perspectives on PII is critical, as their understanding of PII can significantly affect their privacy decisions and practices. While much research has explored users’ privacy perceptions and disclosure preferences regarding PII, less attention has been focused on how users internally define and conceptualize PII. In this study, we conducted interviews with 32 participants to investigate their conceptualization and understanding of PII, using period and fertility tracking apps as the context. Our findings reveal how users perceive the processes and contexts through which personal information, by becoming identifiable, transitions into PII, as well as concerns about data sharing and misuse in these apps. We conclude by advocating for addressing the misalignment between users' perceptions of PII and the regulatory protections and privacy designs surrounding it.

The increasing use of smartphones for capturing documents in various real-world conditions has underscored the need for robust document localization technologies. Current challenges in this domain include handling diverse document types, complex backgrounds, and varying photographic conditions such as low contrast and occlusion. However, there currently are no publicly available datasets containing these complex scenarios and few methods demonstrate their capabilities on these complex scenes. To address these issues, we create a new comprehensive real-world document localization benchmark dataset which contains the complex scenarios mentioned above and propose a novel Real-world Document Localization Network (RDLNet) for locating targeted documents in the wild. The RDLNet consists of an innovative light-SAM encoder and a masked attention decoder. Utilizing light-SAM encoder, the RDLNet transfers the mighty generalization capability of SAM to the document localization task. In the decoding stage, the RDLNet exploits the masked attention and object query method to efficiently output the triple-branch predictions consisting of corner point coordinates, instance-level segmentation area and categories of different documents without extra post-processing. We compare the performance of RDLNet with other state-of-the-art approaches for real-world document localization on multiple benchmarks, the results of which reveal that the RDLNet remarkably outperforms contemporary methods, demonstrating its superiority in terms of both accuracy and practicability.

Few-shot object detection (FSOD) aims to extract semantic knowledge from limited object instances of novel categories within a target domain. Recent advances in FSOD focus on fine-tuning the base model based on a few objects via meta-learning or data augmentation. Despite their success, the majority of them are grounded with parametric readjustment to generalize on novel objects, which face considerable challenges in Industry 5.0, such as 1) a certain amount of fine-tuning time is required and 2) the parameters of the constructed model being unavailable due to the privilege protection, making the fine-tuning fail. Such constraints naturally limit its application in scenarios with real-time configuration requirements or within black-box settings. To tackle the challenges mentioned above, we formalize a novel FSOD task, referred to as test-time few-shot detection (TIDE), where the model is un-tuned in the configuration procedure. To that end, we introduce an asymmetric architecture for learning a support-instance-guided dynamic category classifier. Further, a cross-attention module and a multiscale resizer are provided to enhance the model performance. Experimental results on multiple FSOD platforms reveal that the proposed TIDE significantly outperforms existing contemporary methods. The implementation codes are available at https://github.com/deku-0621/TIDE.

Live streaming has become a popular activity world-wide that has warranted research attention on its privacy related issues. For instance, bystanders' privacy, or the privacy of third-parties captured by streamers, has been increasingly studied as live streaming has become almost ubiquitous in both public and private spaces in many countries. While prior work has studied bystanders' privacy concerns, a gap exists in understanding how streamers consider bystanders' privacy and the steps they take (or do not take) to preserve it. Understanding streamers' considerations towards bystanders' privacy is vital because streamers are the ones who have direct control over whether and how bystanders' information is disclosed. To address this gap, we conducted an interview study with 25 Chinese streamers to understand their considerations and practices regarding bystanders' privacy in live streaming. We found that streamers cared about bystanders' privacy and evaluated possible privacy violations to bystanders from several perspectives. To protect bystanders from privacy violations, streamers primarily relied on technical, behavioral, and collaborative strategies. Our results also indicated that current streaming platforms lacked features that helped streamers seamlessly manage bystanders' privacy and involved bystanders into their privacy decision-making. Applying the theoretical lens of collective privacy management, we discuss implications for the design of live streaming systems to support streamers in protecting bystanders' privacy.

Live streaming is a popular synchronous social media platform that allows users to disclose information to vast audience in real time. It has been increasingly studied in recent years for its unique functions of disseminating user-generated content, enriching streamers' self-presentation, curating online social interactions and fostering online communities. However, little research has been done to explore the privacy issues in live streaming. In the present paper, we aim to understand streamers' privacy concerns and strategies in their information disclosure on live streaming. From an interview study with 20 streamers, we found that they expressed concerns and carefully managed their information disclosure based on whether the disclosure would enhance or weaken their attractiveness to the audience and whether it would disturb their interpersonal boundary with the audience. They adopted various technical and behavioral privacy management strategies to cope with their concerns, but encountered a series of usability and cognitive burdens. Based on the findings, we present design implications to improve the privacy management on live streaming.

2020's COVID-19 crisis has given rise to ubiquitous emergency remote learning (ERL). Guardians, mostly parents, have had to help their children transition and adapt to this difficult learning context. Previous work on remote learning has explored guardian involvement in pre-planned and well-developed remote learning programs, such as established virtual schools. However, ERL lacks pre-planned procedures, policies, and resources. In this paper, we look at how teachers and guardians collaborated to manage the situation. We present an interview study of guardians and teachers of K-12 students in China and look at their collaboration during the COVID-19 ERL. We report how teachers and guardians collaborated to carry out techno-procedural, surveillance, and material work to make ERL possible for K-12 students. Lastly, we reflect on the challenges our participants faced and discuss design implications not only for remote learning during COVID-19 but also future emergency remote learning situations.