Object detection is a key perception technology for autonomous driving, but its performance degrades sharply in fog due to visual signal degradation, which causes reduced contrast and color distortion correlated to scene depth. To address this challenge, this paper proposes the Physics-guided Dual-Branch RT-DETR (PG-DB-RTDETR) framework, a novel teacher-student paradigm. We design a Physics-informed Cross-Attention Fusion (PI-CAF) module that integrates physical principles from the Atmospheric Scattering Model (ASM) into feature fusion. Specifically, the ASM's transmission map is used as a physics-guided spatial gate for adaptive feature compensation, which also endows the model with interpretability by visualizing the physical prior's guidance. To resolve the training-inference inconsistency inherent in teacher-student models, we introduce a lightweight adapter, trained via imitation learning, to ensure real-time operation (55 FPS) with minimal parameter increase. Experiments on Foggy Cityscapes show that the framework provides consistent improvements over the baseline, particularly for small and medium objects, increasing the average precision for small objects (APS) from 5.2% to 9.3%.

Advancements in AI have increased the demand for specialized AI accelerators, with design for general matrix multiplication (GEMM) module being crucial but time-consuming. While large language models (LLMs) show promise for automating GEMM design, challenges arise from GEMM's vast design space and performance requirements. Existing LLM-based frameworks for RTL code generation often lack flexibility and performance awareness. To overcome the challenges, we propose LLM4GV, a multi-agent LLM-based framework that integrates hardware optimization techniques (HOTs) and performance modeling, improving correctness and performance of the generated code over prior works.

Knowledge base question answering is a challenging task that aims to answer questions by querying knowledge bases. Recently, state-of-the-art methods tend to include an enumerator module and a ranker module. They first enumerate by searching the knowledge base and then rank the candidates to select the target logical form. However, these methods sometimes fail to cover the candidates which involve more complex combinations. A recent solution to this issue is to add a generator module after the ranker to generate the uncovered target logical form. However, the enumerator and ranker always discard partial ground truth schema items. Consequently, the lack of them in the generator input results in the failure to generate the target logical form. To address this problem, we present a novel framework, SIMQA, to reuse the neglected schema items, i.e., classes and relations. Specifically, we adopt a matcher module to select the most related schema items for the given question, and feed them to the generator. On this basis, we propose a novel generation model based on contrastive learning to force the model to focus on the supplemental schema items. Experiment results on GRAILQA and WEBQSP datasets demonstrate the highly competitive performance of the proposed method, and verify that schema item matters in KBQA.

Machine reading comprehension (MRC) methods have shown great success in many datasets, but existing methods fail to achieve satisfactory results in low-resource scenarios.In addition, existing MRC models suffer from a notable decrease in performance when confronted with scenes different from the training data.Thus, it is hard to transfer knowledge between domains.In this paper, we propose an adaptive metalearning framework to learn and transfer the shared knowledge.The framework is based on model-agnostic meta-learning algorithm, aiming to aggregate meta-knowledge among multi-source datasets.Furthermore, for better adaptation to different target domain, we investigate an adaptive adversarial training strategy to obtain domain-specific meta-knowledge.We empirically adopt three large-size datasets as source domains and five small-size datasets as target domains, and extensive experiments show the effectiveness of our framework.

In wireless coded cache network, data contents are cached in a number of mobile devices using an erasure correcting code, and a user retrieves content from other mobile devices using device-to-device communication. In this paper, we consider the repair problem when multiple devices that cache data contents fail or leave the network. By exploiting the wireless broadcast nature, we formulate the repair problem over the broadcast channels using an integer linear programming formulation, aiming at minimizing the number of necessary broadcast transmissions. We also study the construction of repair codes and propose a decentralized repair coding method. Simulation results show that the performance using our method outperforms the basic cooperative repair scheme for wired distributed storage systems.

Broadcasting data to multiple users is widely used in wireless applications. We consider a group of mobile users, within proximity of each other, who are interested in the same video content. Network coded broadcast and cooperative coded communication can improve transmission efficiency and throughput over wireless network separately. In this paper we consider the D2D-assisted wireless network coded broadcast problem for users with multiple interfaces to minimize the number of transmission slots. In order to obtain all needed packets, user can receive encoded packet according to cellular link and local cooperative D2D links simultaneously. We analyze the lower bound of number of transmission slots and formulate the problem with integer linear programming(ILP). We also develop heuristic solution for this setup, and simulation results show that our coding scheme significantly reduces the number of transmission slots.

Three RFID reader based network deployment algorithms (grid-covering, diagonal and mixed) were evaluated in this paper. Experimental results show that the grid-covering method can be used to minimize hardware costs, but it leads to many indeterminate positions. The diagonal method can be used to solve the indeterminate problem, however increases the number of readers, especially in a large tracking field. The mixed algorithm can be used to avoid the indeterminate issue and also has the minimum reader number when deployed in a large space. However, it is not suitable for a small tracking field. An optimal deployment algorithm is selected from these three algorithms according to the environmental conditions and the localization requirement. In addition, an optimal RFID reader network deployment combined with a subarea-mapping algorithm can be used to minimize the hardware costs while improving the fine-grained indoor localization accuracy.

Facebook's web site and mobile apps are very dynamic. Every day, they undergo thousands of online configuration changes, and execute trillions of configuration checks to personalize the product features experienced by hundreds of million of daily active users. For example, configuration changes help manage the rollouts of new product features, perform A/B testing experiments on mobile devices to identify the best echo-canceling parameters for VoIP, rebalance the load across global regions, and deploy the latest machine learning models to improve News Feed ranking. This paper gives a comprehensive description of the use cases, design, implementation, and usage statistics of a suite of tools that manage Facebook's configuration end-to-end, including the frontend products, backend systems, and mobile apps.

The information of the distribution of occupant in an indoor environment is important for building energy saving under normal conditions and for evacuation under emergent conditions, and thus is of great practical interest. Due to low set-up cost, wireless sensor networks powered by batteries are usually used for such estimation. The question is how to activate the sensors to minimize the estimation error within a given period of time. In this paper we develop an event-based activation policy which can be easily implemented in a decentralized way. We use numerical experiments to compare this policy with four other policies and to demonstrate the impact of various factors on the performance of these policies, such as the topology, sensor accuracy, battery capacity, occupant movement model, and occupant population. Our method outperforms the other policies in all the tested scenarios.