Recent advances in vision–language models (VLMs) have shed light on human-level embodied intelligence. However, existing benchmarks for VLM-driven embodied agents still rely on high-level commands or discretised action spaces—``non-native'' settings that diverge markedly from the real world. Moreover, current benchmarks focus exclusively on high-level tasks, while lacking joint evaluation and analysis on both low- and high-level. To bridge these gaps, we present \textbf{NativeEmbodied}, a challenging benchmark for VLM-driven embodied agents that adopts a unified, native low-level action space. Built upon diverse simulated scenes, NativeEmbodied first designs three representative high-level tasks in complex scenarios to evaluate overall performance. For more detailed and comprehensive performance analysis, we further decouple the entangled skills behind complex tasks and construct four types of low-level tasks, each corresponding to a key fundamental embodied skill. This joint evaluation across task and skill granularities enables a fine-grained assessment of embodied agent. Comprehensive experiments on the best VLMs reveal pronounced deficiencies in certain fundamental embodied skills. Further analysis shows that these bottlenecks severely constrain performance on high-level tasks. Our NativeEmbodied not only pinpoints the key challenges faced by current VLM-driven embodied agents, but also provides valuable insight for future development of this field.

Communication barriers for the hearing-impaired represent a significant societal concern. Existing sign language recognition solutions are constrained by lighting sensitivity, wearable device burdens, and limited vocabulary coverage. This work presents the first FMCW mmWave radar-based continuous sign language recognition framework, with three key innovations: (1) the Radar Continuous Chinese Sign Language-108 (RCCSL-108) dataset with 108 isolated signs and 66 natural sentences to address the critical absence of sentence-level radar data; (2) a novel lightweight lightweight Continuous Sign Language Recognition Temporal Convolutional Network (CSLR-TCN) that incorporates dual-dilated convolutions for variable-length inputs, achieving 92.66% frame accuracy and robust cross-user generalization; and (3) a novel contextual reordering module that enhances semantic understanding by 6.33% on unseen data by mitigating emotion loss in translation. The implemented embedded system achieves 87.49% recognition accuracy in untrained user experiments with ≥15 FPS throughput, marking a pivotal advancement toward practical radar-based sign language interfaces.

Human activity recognition (HAR) has attracted growing interest due to its wide ranging applications in healthcare, security, and smart surveillance. In particular, radar-based HAR offers a robust, non-contact alternative to conventional sensor based methods. Despite its promise, existing approaches often rely on substantial labeled data and exhibit limited adaptability in small-shot incremental learning scenarios. To address these challenges, we propose a novel few shot class incremental learning (FSCIL) framework for radar-based HAR. Our framework exploits vision transformers (ViTs) augmented with adaptive prompt mechanisms, hybrid attention blocks (HAB), and dynamic feature refinement strategies, which can substantially improve incremental learning performance. We validate the proposed approach on a fused radar dataset containing four sets of diverse radar datasets collected under varying environmental conditions. Our validation shows a harmonic accuracy (HAcc) up to 69.54 % and a Top-1 average accuracy (Top-1 Avg.) up to 91.20 %, outperforming state-of-the-art methods, with a session performance degradation as low as 10.80%. These results underscore the robustness of our approach in recognizing novel activities while preserving previously acquired knowledge.

The task of temporally language grounding (TLG), aiming to locate a video moment within an untrimmed video that matches a given textual query, has attracted considerable research attention in recent years. Typical retrieval-based TLG methods are inefficient due to their reliance on a large number of pre-segmented candidate moments, while localization-based TLG solutions adopt reinforcement learning, resulting in unstable convergence. Meanwhile, the cutting-edge capabilities of multi-modal architecture, especially pre-training paradigm, have not been fully exploited. Therefore, how to perform TLG task efficiently and stably is a non-trivial task. In this work, we propose a novel TLG solution named Multi-modal Multi-Prompt Tuning (MMPT), which formulates the TLG task as a prompt-based multi-modal problem and integrates multiple sub-tasks to tune the performance. In this way, off-the-shelf pre-trained models can be directly leveraged to achieve more stable performance. Specifically, a flexible multi-prompt strategy is contributed to rewrite the query firstly, which contains the query, the start and end timestamps. Among them, various prompt templates are integrated to enhance robustness. Thereafter, a multi-modal Transformer is adopted to fully learn the multi-modal context. Moreover, we design various sub-tasks to optimize this novel framework including the matching task, localization task and joint learning task. Extensive experiments on two real-world datasets validate the effectiveness and rationality of our proposed solution.

Given a textual sentence provided by a user, the Temporal Language Grounding (TLG) task is defined as the process of finding a semantically relevant video moment or clip from an untrimmed video. In recent years, localization-based TLG methods have been explored, which adopt reinforcement learning to locate a clip from the video. However, these methods are not stable enough due to the stochastic exploration mechanism of reinforcement learning, which is sensitive to the reward. Therefore, providing a more flexible and reasonable reward has become a focus of attention for both academia and industry.