In recent years, Gaussian scene representations have achieved a series of promising results in 3D reconstruction. Compared to the previous 3DGS paradigm, the latest reconstruction approach 2DGS can achieve more accurate geometric representation using fewer Gaussian points. Accordingly, developing a panoramic segmentation algorithm suitable for 2DGS-reconstructed scenes is of significant importance. However, existing segmentation methods are primarily designed for 3DGS. They either fail to account for all objects in complex segmentation scenes or suffer from significant performance degradation when applied to 2D Gaussian scenes. Moreover, these methods consistently exhibit poor cross-dataset generalization. To address these issues, we propose IQGS, a segmentation framework applicable to 2DGS representations. Specifically, IQGS employs per-instance query and relaxed object-level supervision instead of strict pixel-level ID supervision, effectively mitigating the segmentation performance degradation that occurs when applied to 2DGS. At the same time, by learning features independent of specific object ID assignments, IQGS enhances its ability to generalize across diverse datasets. Our method achieves impressive panoramic segmentation results across multiple datasets, with an average mIoU of 66.6%, surpassing the state-of-the-art method Gaussian Grouping, which achieves 57.17%.

By providing a standardized interface for LLM agents to interact with external tools, the Model Context Protocol (MCP) is quickly becoming a cornerstone of the modern autonomous agent ecosystem. However, it creates novel attack surfaces due to untrusted external tools. While prior work has focused on attacks injected through external tool outputs, we investigate a more fundamental vulnerability: Tool Poisoning, where malicious instructions are embedded within a tool's metadata at the registration stage. To date, this threat has been primarily demonstrated through isolated cases, lacking a systematic, large-scale evaluation. We introduce MCPTox, the first benchmark to systematically evaluate agent robustness against Tool Poisoning in realistic MCP settings. MCPTox is constructed upon 45 live, real-world MCP servers and 353 authentic tools. To achieve this, we design three distinct attack templates to generate a comprehensive suite of 1348 malicious test cases by few-shot learning, covering 10 categories of potential risks. Our evaluation on 20 prominent LLM agents setting reveals a widespread vulnerability to Tool Poisoning, with GPT-o1-mini, achieving an attack success rate of 72.8%. We find that more capable models are often more susceptible, as the attack exploits their superior instruction-following abilities. Finally, the failure case analysis reveals that agents rarely refuse these attacks, with the highest refused rate (Claude-3.7-Sonnet) less than 3%, demonstrating that existing safety alignment is ineffective against malicious actions that use legitimate tools for unauthorized operation. Our findings create a crucial empirical baseline for understanding and mitigating this widespread threat, and we release MCPTox for the development of verifiably safer AI agents.

Tremor is a common symptom of neurological diseases. The regular assessment of daily tremors facilitates the evaluation of disease progression and assists clinicians in optimizing treatment strategies. However, current home monitoring solutions have difficulty in dealing with user cooperation, privacy concerns, environmental interference, and system generalization, leading to feasibility concerns in activities of daily living (ADL). To this end, we propose Tremerity-Fi, a non-contact and privacy-friendly tremor severity assessment system based on mmWave radar. To realize Tremerity-Fi, we first design an adaptive beamforming algorithm to accurately identify useful but weak signals from numerous reflections captured in the environment. Second, unlike primary reflections commonly used in mmWave sensing, we leverage multipath reflections that carry useful information about the target's motion, even though they are generally considered harmful, to help reconstruct hand signals and improve sensing performance. Furthermore, we propose an unsupervised domain adaptation algorithm to improve the ability to adapt to unseen environments and users. We collect a diverse dataset of 5 patients and 25 healthy subjects in 3 scenarios, such as offices, homes, and hospitals. Extensive experiments show that our system achieves 94.51% accuracy in tremor detection, about 5 higher than the SOTA mmWave radar method, and 89.13% in tremor severity assessment, demonstrating its sufficient potential as a tremor monitoring assistant for patients with neurological diseases.

LLM-powered agents are evolving from passive recommenders into autonomous executors, leveraging tools via the Model Context Protocol (MCP) for web automation. However, this paradigm introduces a new vulnerability: tool poisoning attacks that manipulate the MCP context can corrupt an agent's reasoning. Existing methods focus on anomaly detection and lack autonomous correction mechanisms, hindering their real-world deployment.

Smart-devices' grip-state has shown great potential to enable various intelligent applications, including virtual keyboard and automatic UI adaption. However, smartphones nowadays often lack the capability to detect the grip-state, resulting in a poor experience of human-computer interaction. To implement an effective and efficient grip-state detection, we need to tackle a number of technical challenges such as effective features extraction and fusion from multimodal data, nonalignment of different modal data, and limited labeled data availability. In this work, to address these challenges, we design a two-stage grip-state detecting system, named WowSense, for high-accuracy, real-time detection of phone's grip-states using IMU (Inertial Measurement Unit) and CS (Capacitivc Screen) data. Our system WowSense consists of the multimodal alignment stage and the grip-state classification stage. In the first stage, we employ a novel augmentation method to capture subtle features from IMU and CS data. Additionally, we utilize contrastive learning to extract consistent information across these two modalities using a large amount of unlabeled data. In the second stage, we design an attention-based classifier to capture complementary information using only a small amount of labeled data. We implement our system in OpenHarmony and our extensive experimental results demonstrate the superiority of our system, which achieves 95 % accuracy with only 40 % of the data labeled on a self-collected dataset and 92.5 % accuracy with a latency of only around 10ms when running in real-time on a phone,

Location awareness in mobile devices, particularly the detection of indoor and outdoor transitions, empowers devices to ascertain their own or user's position and offer pertinent intelligent services accordingly. In this paper, we present InOut, a robust and realistic indoor-outdoor detection system characterized by high precision, low latency, and cross-device transferability. Regarding effectiveness, considering that a singular sensor signal is inadequate in providing comprehensive environmental information for detection, we employ a multimodal fusion approach. Concerning efficiency, we optimize the model by pruning non-essential features through the calculation of Shapley values for importance assessment. Furthermore, given the heterogeneity of data from different devices, we implement an unsupervised domain adaptation method that enables effective model transfer across devices with limited unlabeled target domain data. Experimental results demonstrate that our InOut system achieves over 96% accuracy on the test dataset, with detection latency consistently maintained to be within 3.1 seconds (including 3 seconds of interface latency and less than 0.1 seconds of inference latency). Moreover, utilizing unlabeled data from a disparate mobile phone model, amounting to one-sixth the size of the original dataset, we enhance the model's accuracy from 83% before transfer to over 91%.

Considering the long charging time and limitation of available charging stations in the traditional battery recharging facilities, the construction of battery swapping station (BSS) has become a new paradigm to satisfy the energy demand timely and sufficiently. Previous solutions lack a joint consideration of battery swapping station establishment and request assignment with the long distance subsidy. In this paper, we study the Station Placement and Assignment (SPA) problem to minimize the overall operation cost. Unfortunately, it shows great difficulty due to the infinite candidate locations and the complex coupling relation for request assignments. To address these challenges, we first devise the bundle generation strategy to reduce the infinite candidate location to finite, then propose an efficient algorithm based on the greedy strategy to assign the battery swapping requests. Extensive evaluations are carried out to show the outstanding performance of our proposed algorithms.