Efficient resume parsing is critical for global hiring, yet the absence of dedicated benchmarks for evaluating large language models (LLMs) on multilingual, structure-rich resumes hinders progress. To address this, we introduce ResumeBench, the first privacy-compliant benchmark comprising 2,500 synthetic resumes spanning 50 templates, 30 career fields, and 5 languages. These resumes are generated through a human-in-the-loop pipeline that prioritizes realism, diversity, and privacy compliance, which are validated against real-world resumes. This paper evaluates 24 state-of-the-art LLMs on ResumeBench, revealing substantial variations in handling resume complexities. Specifically, top-performing models like GPT-4o exhibit challenges in cross-lingual structural alignment while smaller models show inconsistent scaling effects. Code-specialized LLMs underperform relative to generalists, while JSON outputs enhance schema compliance but fail to address semantic ambiguities. Our findings underscore the necessity for domain-specific optimization and hybrid training strategies to enhance structural and contextual reasoning in LLMs.

Substitute training for black-box attacks, leveraging query synthesis, has recently gained significant attention, especially in data-free scenarios. Existing works focus on extracting the victim model without exploring the deployment of adversarial attacks, leading to inefficiencies and redundant queries. To address these issues, we present a novel data-free substitute training attack framework that focuses on aligning the ’chain of attack’ rather than the decision boundary of models. Specifically, we propose a Temporal Contrastive Sampling module to capture essential discriminative features. A Simplified Adversarial Training module is further introduced to restrict the routes crossing the decision boundary. For catastrophic forgetting and data imbalance considerations, we integrate a Substitute Model Resetting module strategy. Comprehensive experiments over three datasets demonstrate our effectiveness: under decision-based settings, we can achieve targeted attack success rates of 81.87% and 91.89% on the CIFAR-10 dataset with query budgets of 500k and 1M, surpassing state-of-the-art methods under score-based settings.

Utilizing unique physiological or behavioral traits, biometrics offers an intuitive authentication approach. However, common biometric modalities are susceptible to ambient factors and privacy concerns. This paper proposes CaphandAuth, a novel capacitive touchscreen-based hand authentication system. Using intrinsic capacitive imaging within the touchscreen, it provides a new secure, cost-effective, and user-friendly biometric authentication solution that is inherently resilient to environmental factors. To this end, CaphandAuth captures consecutive capacitive frames as the hand moves across the touchscreen. These frames are processed with an innovative super-resolution algorithm tailored for deformable objects to enhance details. A learning-based feature extractor then derives expressive and adaptive feature representations from the enhanced images. Extensive experiments demonstrate that CaphandAuth achieves an authentication accuracy of 99.84% and an equal error rate (EER) of 2.77% on a commercial tablet. Moreover, Caphand-Auth exhibits formidable resilience to diverse deceiving attempts, including handprint simulation attacks, counterfeit spoofing attacks, and puppet attacks, making it a robust and secure solution in real-world scenarios.

When facing changing environments in the real world, the lightweight model on client devices suffers from severe performance drops under distribution shifts. The main limitations of the existing device model lie in (1) unable to update due to the computation limit of the device, (2) the limited generalization ability of the lightweight model. Meanwhile, recent large models have shown strong generalization capability on the cloud while they can not be deployed on client devices due to poor computation constraints. To enable the device model to deal with changing environments, we propose a new learning paradigm of Cloud-Device Collaborative Continual Adaptation, which encourages collaboration between cloud and device and improves the generalization of the device model. Based on this paradigm, we further propose an Uncertainty-based Visual Prompt Adapted (U-VPA) teacher-student model to transfer the generalization capability of the large model on the cloud to the device model. Specifically, we first design the Uncertainty Guided Sampling (UGS) to screen out challenging data continuously and transmit the most out-of-distribution samples from the device to the cloud. Then we propose a Visual Prompt Learning Strategy with Uncertainty guided updating (VPLU) to specifically deal with the selected samples with more distribution shifts. We transmit the visual prompts to the device and concatenate them with the incoming data to pull the device testing distribution closer to the cloud training distribution. We conduct extensive experiments on two object detection datasets with continually changing environments. Our proposed U-VPA teacher-student framework outperforms previous state-of-the-art test time adaptation and device-cloud collaboration methods. The code and datasets will be released.