Large Language Models (LLMs), renowned for their superior proficiency in language comprehension and generation, stimulate a vibrant ecosystem of applications around them. However, extensive assimilation into various services introduces significant security risks. This study deconstructs the complexities implications prompt injection attacks on actual LLM-integrated applications. Initially, we conduct an exploratory analysis ten commercial applications, highlighting constraints current attack...

With the rapid development and large-scale popularity of program software, modern society increasingly relies on software systems. However, problems exposed by have also come to fore. Software defect has become an important factor troubling developers. In this context, Automated Program Repair (APR) techniques emerged, aiming automatically fix reduce manual debugging work. particular, benefiting from advances in deep learning, numerous learning-based APR emerged recent years, which bring new...

Bitmap indexes are widely used in both scientific and commercial databases. They bring fast read performance for specific types of queries, such as equality selective range queries. A major drawback bitmap indexes, however, is that supporting updates particularly costly. kept compressed to minimize storage footprint; a result, updating index requires the expensive step decoding then encoding bitvector. Today, more applications need support reads writes, blurring boundaries between analytical...

Retrieval finds a small number of relevant candidates from large corpus for information retrieval and recommendation applications. A key component is to model (user, item) similarity, which commonly represented as the dot product two learned embeddings. This formulation permits efficient inference, known Maximum Inner Product Search (MIPS). Despite its popularity, products cannot capture complex user-item interactions, are multifaceted likely high rank. We hence examine non-dot-product...

All-Multi-Layer Perceptron (all-MLP) mixer models have been shown to be effective for time series forecasting problems. However, when such a model is applied high-dimensional (e.g., the in spatial-temporal dataset), its performance likely degrade due overfitting issues. In this paper, we propose an all-MLP architecture, referred as RPMixer. Our method leverages ensemble-like behavior of deep neural networks, where each individual block within network acts like base learner ensemble model,...

The state space in Multiagent Reinforcement Learning (MARL) grows exponentially with the agent number. Such a curse of dimensionality results poor scalability and low sample efficiency, inhibiting MARL for decades. To break this curse, we propose unified permutation framework that exploits invariance (PI) equivariance (PE) inductive biases to reduce multiagent space. Our insight is permuting order entities factored does not change information. Specifically, two novel implementations: Dynamic...

Software-Defined Networking (SDN) has gained special attention in both academia and industry. It is a new network architecture framework for networking, which decouples the control plane from data at physical topology. SDN promotes centralization of introduces t

Despite many breakthroughs in recent years, it is still hard for MultiAgent Reinforcement Learning (MARL) algorithms to directly solve complex tasks Systems (MASs) from scratch. In this work, we study how use Automatic Curriculum (ACL) reduce the number of environmental interactions required learn a good policy. order difficult task, ACL methods automatically select sequence (i.e., curricula). The idea obtain maximum learning progress towards final task by continuously on that match current...

Numerous algorithms have been developed for online product rating prediction, but the specific influence of user and information in determining final prediction score remains largely unexplored. Existing research often relies on narrowly defined data settings, which overlooks real-world challenges such as cold-start problem, cross-category utilization, scalability deployment issues. To delve deeper into these aspects, particularly to uncover roles individual taste collective wisdom, we...

Self-supervised Pretrained Models (PTMs) have demonstrated remarkable performance in computer vision and natural language processing tasks. These successes prompted researchers to design PTMs for time series data. In our experiments, most self-supervised were surpassed by simple supervised models. We hypothesize this undesired phenomenon may be caused data scarcity. response, we test six generation methods, use the generated pretraining lieu of real data, examine effects on classification...

Binary code similarity detection (BinSD) systems trend to utilize deep learning identify semantic features of assembly and exhibits superior performance, gaining increasing popularity against traditional methods. However, it has been observed that existing models are susceptible data poisoning attacks, posing a latent threat the robustness reliability BinSD. Existing strategies in BinSD easily detectable for generated triggers will destroy functions. Moreover, selecting trigger injection...

The title compound (systematic name: 9α-fluoro-11β,17α,21-tri­hydroxy-16β-methyl-3,20-dioxopregna-1,4-diene-17,21-diyl dipropionate, BMDP), C28H37FO7, crystallizes in the ortho­rhombic space group P212121. overall shape of steroid skeleton is similar to that other anti-inflammatory steroids. Ring A planar due 1,4-dien-3-one moiety, rings B and C have normal chair conformations ring D adopts an envelope conformation.

Abstract In the smart manufacturing factory, fresh real-time datas and multiple replications are important for numerically-controlled machine tools. The production stored in distributed cloud data centers (CDCs), so communication latencies produced by synchronization may impair machining performance. this paper, an asynchronous consistency algorithm (ACA) which is inspired nearest infection proposed to reduce between nodes. experimental results demonstrate that can support lower client-side...

Abstract Background: Management of pharmaceuticals and personal care products (PPCPs) in the environment has become a social issue. In present study, concentrations 140 PPCPs at 20 sites Baiyangdian Lake Tai from 2016 to 2017 were analyzed by Ultra Performance Liquid Chromatography Mass Spectrometer (UPLC-MS). Risk quotients (RQ) calculated for each detected chemical all prioritization indices (PI), based on maximum RQ, calculated. To assess risk chemicals that identified high priority...

Word vector embeddings have been shown to contain and amplify biases in data they are extracted from. Consequently, many techniques proposed identify, mitigate, attenuate these word representations. In this paper, we utilize interactive visualization increase the interpretability accessibility of a collection state-of-the-art debiasing techniques. To aid this, present Visualization Embedding Representations for deBiasing system ("VERB"), an open-source web-based tool that helps users gain...

Against the backdrop of an increasingly serious global ecological crisis, scholars have turned their attention to Eastern Confucianism in attempt find wisdom traditional order seek a path harmonious development between human beings and nature. Kumazawa Banzan, as Confucianist realist with environmental consciousness who emerged context severe destruction early Edo period, is very representative. In this paper, we will take Kumazawa’s object study explore his Confucianism, taking mountain...

Pre-training on large models is prevalent and emerging with the ever-growing user-generated content in many machine learning application categories. It has been recognized that contextual knowledge from datasets depicting user-content interaction plays a vital role downstream tasks. Despite several studies attempting to learn via pre-training methods, finding an optimal training objective strategy for this type of task remains challenging problem. In work, we contend there are two distinct...

In recent years, the building materials industry in China has made great progress R&D of energy conservation, emission reduction and cleaner production technologies, order to implement sustainable development policy. Life cycle assessment (LCA) is one mainstream method analyze environmental impact product during its life cycle, which plays an important role on ecological design green manufacture technology year. This paper reviewed LCA studies materials. Firstly, China's technical...

Plug-in hybrid electric vehicles (PHEVs) are widely used in the field of new energy due to their improvement power efficiency and low emission. This paper proposes a cloud-edge monitoring system (CEMS) architecture for PHEV, which includes container as service (CaaS) cloud platform edge nodes realize PHEVs status data storage analysis. In CaaS, open-source software featured proposed enables user implement processing. Then scheduling algorithm is optimize querying speed. Finally, simulation...