Despite the integration of safety alignment and external filters, text-to-image (T2I) generative systems are still susceptible to producing harmful content, such as sexual or violent imagery. This raises serious concerns about unintended exposure and potential misuse. Red teaming, which aims to proactively identify diverse prompts that can elicit unsafe outputs from the T2I system, is increasingly recognized as an essential method for assessing and improving safety before real-world deployment. However, existing automated red teaming approaches often treat prompt discovery as an isolated, prompt-level optimization task, which limits their scalability, diversity, and overall effectiveness. To bridge this gap, in this paper, we propose DREAM, a scalable red teaming framework to automatically uncover diverse problematic prompts from a given T2I system. Unlike prior work that optimizes prompts individually, DREAM directly models the probabilistic distribution of the target system's problematic prompts, which enables explicit optimization over both effectiveness and diversity, and allows efficient large-scale sampling after training. To achieve this without direct access to representative training samples, we draw inspiration from energy-based models and reformulate the objective into a simple and tractable form. We further introduce GC-SPSA, an efficient optimization algorithm that provides stable gradient estimates through the long and potentially non-differentiable T2I pipeline. During inference, we also propose a diversity-aware sampling strategy to enhance prompt variety. The effectiveness of DREAM is validated through extensive experiments, demonstrating state-of-the-art performance across a wide range of T2I models and safety filters in terms of both prompt success rate and diversity. Our code is available at https://github.com/AntigoneRandy/DREAM

Recently, on-device large models (e.g., 7B parameter LLMs and LVLMs) are increasingly deployed in real-world services such as mobile assistants, edge computing, and IoT systems, where low latency and resource efficiency are critical. Since large models trained on general-domain data are typically suboptimal for downstream services, In-Context Learning (ICL) technique is widely used to enhance their task-specific service performance without extra tuning. However, effective ICL methods generally necessitate a sufficient amount of supervised data to provide abundant service-related information, which usually leads to a long context (i.e., more texts or images input) for models to process. This long context issue can significantly degrade the real-time inference efficiency and also impact the information utilization effectiveness, particularly for on-device large models with limited long context modeling capacities. To tackle this challenge, we propose a Personalized Knowledge Refinement (PKR) framework to achieve efficient and effective ICL for on-device large models. Specifically, we first introduce a personalized knowledge extraction module, which analyzes the inference behaviors of the target model on small-scale supervised data and then convert these behavioral patterns into personalized service-specific knowledge as the instruction context. Furthermore, we propose an adaptive knowledge filtration mechanism to model the informativeness of the extracted knowledge and eliminate the redundant ones, further improving the knowledge encoding efficiency per unit of context length. Experiments based on 14 benchmarks spanning both textual and visual task services, and 4 large models, demonstrated that PKR consistently improves task accuracy while reducing context length and inference latency, making it a practical solution for real-world on-device services. Codes are released athttps://github.com/wanghl21/PKR.

Large vision-language models (LVLMs) derive their capabilities from extensive training on vast corpora of visual and textual data. Empowered by large-scale parameters, these models often exhibit strong memorization of their training data, rendering them susceptible to membership inference attacks (MIAs). Existing MIA methods for LVLMs typically operate under white- or gray-box assumptions, by extracting likelihood-based features for the suspected data samples based on the target LVLMs. However, mainstream LVLMs generally only expose generated outputs while concealing internal computational features during inference, limiting the applicability of these methods. In this work, we propose the first black-box MIA framework for LVLMs, based on a prior knowledge-calibrated memory probing mechanism. The core idea is to assess the model memorization of the private semantic information embedded within the suspected image data, which is unlikely to be inferred from general world knowledge alone. We conducted extensive experiments across four LVLMs and three datasets. Empirical results demonstrate that our method effectively identifies training data of LVLMs in a purely black-box setting and even achieves performance comparable to gray-box and white-box methods. Further analysis reveals the robustness of our method against potential adversarial manipulations, and the effectiveness of the methodology designs. Our code and data are available at \url{https://github.com/spmede/KCMP}.

In recent years, antimicrobial peptides (AMPs) have attracted interest as potential peptide antibiotic due to their broad-spectrum antibacterial activity and high target specificity. However, existing research on AMP prediction mainly focuses on their functional properties, such as antibacterial, antiviral, and anticancer. This emphasis has created a significant gap in identifying AMPs that specifically target pathogens. Given the large variety of pathogens and the sparsity and imbalance of labels, it is challenging to determine which specific pathogens AMPs can effective against. To address this issue, we present PepXML, a large language model-based tool for extreme multilabel classification of pathogen-targeted AMPs. Our first step involved constructing a benchmark dataset of AMPs and their corresponding targeted pathogens, sourced from public databases. In PepXML, the peptides are embedded using ESM2. Further, clustering on a specifically designed label co-occurrence graph and hard negative sampling were employed to address challenges on data sparsity and label imbalance. To validate the reliability of our predictive results, we conducted molecular docking studies focused on peptide-bilayer membrane interactions and performed molecular dynamics simulations to elucidate the mechanisms of peptide-pathogen interactions. We anticipate that PepXML will be a valuable resource for advancing peptide-based therapeutics. The data and Python codes of the PepXML model are available at https://github.com/YannanBin/PepXML.git.

The inappropriate use of antibiotics has precipitated the emergence of multidrug-resistant bacteria, prompting significant interest in antimicrobial peptides (AMPs) as potential alternatives to traditional antibiotics. Given the prohibitive costs and time-consuming nature of biological experiments, computational methods provide an efficient alternative for the development of AMP-based drugs. However, existing computational studies primarily focus on identifying AMPs with antimicrobial activity, lacking a targeted identification of AMPs against specific microbial species. To address this gap, we propose a peptide-microbe association (PMA) prediction framework, termed AEPMA, which is constructed based on an autoevolutionary heterogeneous graph. Within AEPMA, we construct an innovative peptide-microbe-disease network (PMDHAN). Furthermore, we design an autoevolutionary information aggregation mechanism that facilitates the representation learning of the heterogeneous graph. This model automatically aggregates semantic information within the heterogeneous network while thoroughly accounting for the spatiotemporal dependencies and heterogeneous interactions in the PMDHAN. Experiments conducted on one peptide-microbe and three drug-microbe association datasets demonstrate that the performance of AEPMA outperforms five state-of-the-art methods, demonstrating its robust modeling capability and exceptional generalization ability. In addition, this study identifies a novel anti-Staphylococcus aureus peptide and an anti-Escherichia coli peptide, thereby contributing valuable information for the development of antimicrobial drugs and strategies for mitigating antibiotic resistance.

Prediction of drug-target interactions (DTIs) is essential in medicine field, since it benefits the identification of molecular structures potentially interacting with drugs and facilitates the discovery and reposition of drugs. Recently, much attention has been attracted to network representation learning to learn rich information from heterogeneous data. Although network representation learning algorithms have achieved success in predicting DTI, several manually designed meta-graphs limit the capability of extracting complex semantic information. To address the problem, we introduce an adaptive meta-graph-based method, termed AMGDTI, for DTI prediction. In the proposed AMGDTI, the semantic information is automatically aggregated from a heterogeneous network by training an adaptive meta-graph, thereby achieving efficient information integration without requiring domain knowledge. The effectiveness of the proposed AMGDTI is verified on two benchmark datasets. Experimental results demonstrate that the AMGDTI method overall outperforms eight state-of-the-art methods in predicting DTI and achieves the accurate identification of novel DTIs. It is also verified that the adaptive meta-graph exhibits flexibility and effectively captures complex fine-grained semantic information, enabling the learning of intricate heterogeneous network topology and the inference of potential drug-target relationship.

With the rapid development of Graph Neural Networks (GNNs), more and more studies focus on system design to improve training efficiency while ignoring the efficiency of GNN inference. Actually, GNN inference is a non-trivial task, especially in industrial scenarios with giant graphs, given three main challenges, i.e., scalability tailored for full-graph inference on huge graphs, inconsistency caused by stochastic acceleration strategies (e.g., sampling), and the serious redundant computation issue. To address the above challenges, we propose a scalable system named InferTurbo to boost the GNN inference tasks in industrial scenarios. Inspired by the philosophy of "think-like-a-vertex", a GAS-like (Gather-Apply-Scatter) schema is proposed to describe the computation paradigm and data flow of GNN inference. The computation of GNNs is expressed in an iteration manner, in which a vertex would gather messages via in-edges and update its state information by forwarding an associated layer of GNNs with those messages and then send the updated information to other vertexes via out-edges. Following the schema, the proposed InferTurbo can be built with alternative backends (e.g., batch processing system or graph computing system). Moreover, InferTurbo introduces several strategies like shadow-nodes and partial-gather to handle nodes with large degrees for better load balancing. With InferTurbo, GNN inference can be hierarchically conducted over the full graph without sampling and redundant computation. Experimental results demonstrate that our system is robust and efficient for inference tasks over graphs containing some hub nodes with many adjacent edges. Meanwhile, the system gains a remarkable performance compared with the traditional inference pipeline, and it can finish a GNN inference task over a graph with tens of billions of nodes and hundreds of billions of edges within 2 hours.

Machine learning over graphs has been emerging as powerful learning tools for graph data. However, it is challenging for industrial communities to leverage the techniques, such as graph neural networks (GNNs), and solve real-world problems at scale because of inherent data dependency in the graphs. As such, we cannot simply train a GNN with classic learning systems, for instance, parameter server that assumes data parallelism. Existing systems store the graph data in-memory for fast accesses either in a single machine or graph stores from remote. The major drawbacks are three-fold. First, they cannot scale because of the limitations on the volume of the memories, or the bandwidth between graph stores and workers. Second, they require extra development of graph stores without well exploiting mature infrastructures such as MapReduce that guarantee good system properties. Third, they focus on training but ignore optimizing the performance of inference over graphs, thus makes them an unintegrated system. In this paper, we design AGL, a scalable and integrated system, with fully-functional training and inference for GNNs. Our system design follows the message passing scheme underlying the computations of GNNs. We design to generate the K -hop neighborhood, an information-complete subgraph for each node, as well as do the inference simply by merging values from in-edge neighbors and propagating values to out-edge neighbors via MapReduce. In addition, the K -hop neighborhood contains information-complete subgraphs for each node, thus we simply do the training on parameter servers due to data independence. Our system AGL, implemented on mature infrastructures, can finish the training of a 2-layer GNN on a graph with billions of nodes and hundred billions of edges in 14 hours, and complete the inference in 1.2 hours.