Understanding and reasoning over 3D scenes from natural language queries is essential for intuitive human-machine interaction in robotics, navigation, and mixed reality. While existing VLM-based methods address this task by training on 3D data, they face significant challenges in scalability and adaptability due to the limited availability of such data. Training-free approaches offer a promising alternative but often depend on fixed camera paths, coarse view sampling, or preprocessed inputs, limiting effective reasoning. We propose TAVEN, a training-free framework that leverages MLLM (i.e., GPT) for adaptive, task-driven exploration of 3D scenes. TAVEN features: 1) Chain-of-Thought Query Decomposition with Dual-Focus Reasoning to break down queries into sub-tasks focused on relevant entities and goals; 2) Global Memory-Based Dual-Level Retrieval for retrieving contextually relevant views; 3) Progressive View Adjustment with Tri-Criteria Evaluation to iteratively refine viewpoints; and 4) Trajectory Aware View Rectification to suggest improved (rectified) views based on exploration history. By integrating visual feedback with task semantics, TAVEN enables zero-shot, goal-oriented 3D reasoning, moving beyond passive perception toward intelligent spatial understanding.

This study explores the dynamics of a gated memory car-following system, with a focus on the challenges encountered when training models using fine-grained spatiotemporal data. To address the issues of redundant gradient updates and limited generalization inherent in traditional sequential training methods, a novel Spatiotemporal-Decoupled Training (SDT) method is proposed. This method enhances gradient variance by decoupling temporal dependencies and mixing trajectory segments from different vehicles, thereby improving model generalization performance and achieving a zero collision rate on test dataset. Experimental validation is carried out using three datasets (HighD, NGSIM-I80 and Lyft) and two basic models (GRU and LSTM) to assess the effectiveness of the proposed method. The results demonstrate significant improvements in model performance, including an 80% reduction in generalization error on the HighD dataset, a 14% reduction on the NGSIM-I80 dataset a 57% reduction on Lyft dataset, and the achievement of a Zero-collision rate on all test datasets, showcasing the potential of the SDT method for intelligent driving systems. Our code and experimental configurations are publicly available on GitHub to facilitate reproducibility and comparison:https://github.com/LiangzgJlu/Spatiotemporal-Decoupled-Training

Recently, with the development of Neural Radiance Fields and Gaussian Splatting, 3D reconstruction techniques have achieved remarkably high fidelity. However, the latent representations learnt by these methods are highly entangled and lack interpretability. In this paper, we propose a novel part-aware compositional reconstruction method, called GaussianBlock, that enables semantically coherent and disentangled representations, allowing for precise and physical editing akin to building blocks, while simultaneously maintaining high fidelity. Our GaussianBlock introduces a hybrid representation that leverages the advantages of both primitives, known for their flexible actionability and editability, and 3D Gaussians, which excel in reconstruction quality. Specifically, we achieve semantically coherent primitives through a novel attention-guided centering loss derived from 2D semantic priors, complemented by a dynamic splitting and fusion strategy. Furthermore, we utilize 3D Gaussians that hybridize with primitives to refine structural details and enhance fidelity. Additionally, a binding inheritance strategy is employed to strengthen and maintain the connection between the two. Our reconstructed scenes are evidenced to be disentangled, compositional, and compact across diverse benchmarks, enabling seamless, direct and precise editing while maintaining high quality.

Vulnerability is a significant security threat to information systems, drawing widespread concern from researchers. In recent years, owing to continuous advancements in defense technologies, the success rate of exploiting a single N-day vulnerability for attacking has gradually decreased. Attackers are now attempting to exploit multiple vulnerabilities simultaneously to achieve their objectives. This phenomenon is referred to as the vulnerability co-exploitation. Limited by strict vulnerability triggering conditions, successful attacks via vulnerability co-exploitation are infrequent. Due to the low proportion of co-exploitation cases among all vulnerabilities, few studies have focused on co-exploitation relationships or investigated co-exploitation under the condition of extreme data imbalance. In additon, existing work lacks sufficient multidimensional features, which are crucial for accurately identifying and understanding co-exploitation scenarios. However, the prediction of vulnerability co-exploitation remains valuable as it aids practitioners in identifying potential critical risk points within the system. In this article, we propose a framework named CoExpMiner, based on the attributed heterogeneous information network, for mining potential vulnerability co-exploitation under the extreme data imbalance condition. CoExpMiner utilizes structure and attribute features of the attributed heterogeneous graph to predict vulnerability co-exploitation, with employing a prefilter structure to accelerate the process and reduce the computational cost. Experimental results demonstrate that CoExpMiner can effectively predict co-exploitation despite the challenges posed by extreme data imbalance.

In recent years, software supply chain attacks have become increasingly prevalent, prompting considerable research into detecting malicious packages within relevant repositories. With the popularity bolstered by the widespread adoption of open-source practices, Java become one of the preferred languages among modern developers. However, the issue of malware detection in Java components remains unresolved. Most prior approaches suffer from insufficient code coverage and coarse-grained representation, making them unsuitable for Java components.In this paper, we propose an innovative solution calledSheartailored for detecting malicious Java components.Shearfirstly analyzes all methods in the component and locates potential malicious code snippets based on sensitive calls, as slice-level analysis provides a better understanding of the specific malicious activities. Secondly, statements depending on sensitive call sites are extracted and embedded into vectors for further detection instead of function-level representation which is coarse-grained facing the dynamic features in Java. The corresponding experimental results show thatSheareffectively identifies the malicious semantics hidden in the code slices by leveraging the neural network model, outperforming currently available tools to a great extent. Through real-world validation,Sheardetected 51 components with malicious characteristics out of 68,273, demonstrating its practical feasibility. This study introduces the first Java malicious component detection method suitable for real-world scenarios, carrying considerable practical significance in bolstering defenses within the software supply chain.

Approximately, 350 million people, a proportion of 8%, suffer from color vision deficiency (CVD). While image generation algorithms have been highly successful in synthesizing high-quality images, CVD populations are unintentionally excluded from target users and have difficulties understanding the generated images as normal viewers do. Although a straightforward baseline can be formed by combining generation models and recolor compensation methods as the post-processing, the CVD friendliness of the result images is still limited since the input image content of recolor methods is not CVD-oriented and will be fixed during the recolor compensation process. Besides, the CVD populations can not be fully served since the varying degrees of CVD are often neglected in recoloring methods. Instead, we propose a personalized CVD-friendly image generation algorithm with two key characteristics: (i) generating CVD-oriented images aligned with the needs of CVD populations; (ii) generating continuous personalized images for people with various CVD degrees through disentangling the color representation based on a triple-latent structure. Quantitative and qualitative experiments indicate our proposed image generation model can generate practical and compelling results compared to the normal generation model and combination baselines on several datasets. The code is available at: https://github.com/Jiangshuyi0V0/CVD-GAN.git