Multi-hop question answering is a fundamental challenge for RAG, as it requires stepwise reasoning and progressive integration of evidence from multiple sources. Most existing approaches decompose complex questions into multiple sub-questions at once and perform retrieval in parallel, which neglects dependencies among sub-questions and frequently results in broken reasoning chains. Furthermore, the use of a single retrieval source—either unstructured text or structured knowledge graphs—cannot adequately support the heterogeneous information requirements of different reasoning steps. In this work, we propose a multi-agent iterative RAG framework that enables dependency-aware reasoning through evidence-driven sub-question generation. A task-planning agent incrementally generates single-hop sub-questions by identifying evidence gaps from an evolving evidence set, ensuring that each sub-question is well-posed and contextually grounded. Meanwhile, the framework dynamically selects between Graph-RAG Agent and Text-RAG Agent at the sub-question level, exploiting their complementary strengths in relational reasoning and factual coverage. Extensive experiments on HotpotQA, 2WikiMultihopQA, and MuSiQue show that our approach consistently outperforms strong baselines, achieving state-of-the-art accuracy while maintaining greater robustness and stability as reasoning depth increases.

Activation steering is an efficient technique for aligning the behavior of large language models (LLMs) by injecting steering vectors directly into a model’s residual stream during inference. A pivotal challenge in this approach lies in choosing the right layers to intervene, as inappropriate selection can undermine behavioral alignment and even impair the model’s language fluency and other core capabilities. While single-layer steering allows straightforward evaluation on held-out data to identify the "best" layer, it offers only limited alignment improvements. Multi-layer steering promises stronger control but faces a combinatorial explosion of possible layer subsets, making exhaustive search impractical. To address these challenges, we propose LayerNavigator, which provides a principled and promising layer selection strategy. The core innovation of LayerNavigator lies in its novel, quantifiable criterion that evaluates each layer's steerability by jointly considering two key aspects: discriminability and consistency. By reusing the activations computed during steering vector generation, LayerNavigator requires no extra data and adds negligible overhead. Comprehensive experiments show that LayerNavigator achieves not only superior alignment but also greater scalability and interpretability compared to existing strategies. Our code is available at https://github.com/Bryson-Arrot/LayerNavigator

Multiple instance learning (MIL) trains models from bags of instances, where each bag contains multiple instances, and only bag-level labels are available for supervision. The application of graph neural networks (GNNs) in capturing intrabag topology effectively improves MIL. Existing GNNs usually require filtering low-confidence edges among instances and adapting graph neural architectures to new bag structures. However, such asynchronous adjustments to structure and architecture are tedious and ignore their correlations. To tackle these issues, we propose a reinforced GNN framework for MIL (RGMIL), pioneering the exploitation of multiagent deep reinforcement learning (MADRL) in MIL tasks. MADRL enables the flexible definition or extension of factors that influence bag graphs or GNNs and provides synchronous control over them. Moreover, MADRL explores structure-to-architecture correlations while automating adjustments. Experimental results on multiple MIL datasets demonstrate that RGMIL achieves the best performance with excellent explainability. The code and data are available at https://github.com/RingBDStack/RGMIL.

Social bots frequently engage in malicious activities like spreading misinformation and phishing on major social media platforms, significantly impacting the fairness and security of these platforms. Therefore, detecting social bots has become a very critical task. However, we observe two challenges for bot detection methods: neglected discrepancies under various interests (e.g., politics, entertainment) and challenging cases in the real world (e.g., carefully camouflaged bots, individualized genuine users). To tackle these issues, we propose BotCHMIA, a novel interest-aware social bot detection method enhanced with challenging cases. Specifically, to enhance feature representations by various user interests, we propose an interest-aware feature collaboration that utilizes a series of expert networks and an interest adapter to acquire user interest-specific information and fuse it with task-specific feature representations extracted by a bot detection projection. Additionally, we estimate sample hardness during the training process based on the model’s classification confidence and improve existing supervised contrastive loss with randomly selected challenging cases, namely hard samples, to enhance the discriminability of user feature representations. We conduct extensive experiments on two real social bot datasets, and the results demonstrate the practical benefits gained from our proposed detection method.

Aspect-based sentiment analysis (ABSA) is dedicated to forecasting the sentiment polarity of aspect terms within sentences. Employing graph neural networks to capture structural patterns from syntactic dependency parsing has been confirmed as an effective approach for boosting ABSA. In most works, the topology of dependency trees or dependency-based attention coefficients is often loosely regarded as edges between aspects and opinions, which can result in insufficient and ambiguous syntactic utilization. To address these problems, we propose a new reinforced dependency graph convolutional network (RDGCN) that improves the importance calculation of dependencies in both distance and type views. Initially, we propose an importance calculation criterion for the minimum distances over dependency trees. Under the criterion, we design a distance-importance function that leverages reinforcement learning for weight distribution search and dissimilarity control. Since dependency types often do not have explicit syntax like tree distances, we use global attention and mask mechanisms to design type-importance functions. Finally, we merge these weights and implement feature aggregation and classification. Comprehensive experiments show the effectiveness of the criterion and importance functions. RDGCN yields excellent analysis results.

Synthetic lethality (SL) prediction is used to identify if the co-mutation of two genes results in cell death. The prevalent strategy is to abstract SL prediction as an edge classification task on gene nodes within SL data and achieve it through graph neural networks (GNNs). However, GNNs suffer from limitations in their message passing mechanisms, including over-smoothing and over-squashing issues. Moreover, harnessing the information of non-SL gene relationships within large-scale multi-omics data to facilitate SL prediction poses a non-trivial challenge. To tackle these issues, we propose a new multi-omics sampling-based graph transformer for SL prediction (MSGT-SL). Concretely, we introduce a shallow multi-view GNN to acquire local structural patterns from both SL and multi-omics data. Further, we input gene features that encode multi-view information into the standard self-attention to capture long-range dependencies. Notably, starting with batch genes from SL data, we adopt parallel random walk sampling across multiple omics gene graphs encompassing them. Such sampling effectively and modestly incorporates genes from omics in a structure-aware manner before using self-attention. We showcase the effectiveness of MSGT-SL on real-world SL tasks, demonstrating the empirical benefits gained from the graph transformer and multi-omics data.

Abstract To analyse the computed tomography (CT) images of spiral neuron T‐type calcification channels and the mechanism of action in patients with sudden deafness in the elderly, the artificial intelligence (edge algorithm) algorithm was applied to the embedded system to analyse the CT images of the sudden deafness of the elderly, and then the mechanism of the spiral neuron T‐calcification channel was studied. The results showed that among the 48 patients with labyrinthitis, 26 were acute labyrinthitis, 14 were chronic labyrinthitis, and 8 were sclerosing labyrinthitis. α1G, α1H, and α1I were expressed on the cochlea and spiral neurons in the 66–68‐year‐old population, but the expression levels were significantly different. The expression of the three receptors on spiral neurons was slightly higher than that of the cochlea. α1H was most expressed on spiral neurons compared with the other two calcium channel receptors (p < 0.05). In addition, the minimum intensity projection method, multi‐layer reconstruction method, and surface reconstruction method can fully display the bony labyrinth, cochlea, cochlear duct (2.5 weeks), and shape and edge damage. In summary, the adoption of artificial intelligence algorithms based on embedded systems to analyse CT images can comprehensively and accurately reflect the three‐dimensional structure of the inner surface of the inner ear labyrinth. It can also fully display the scope, shape, density, and edge of the lesion. T‐type calcium channel receptors were expressed on the cochlea and spiral neurons of the C57BL/6J elderly, and decreased with age.

Event detection on social platforms can help people perceive essential events and make actionable decisions. Existing document-pivot streaming social event detection methods generally embed documents and perform text clustering. They face the challenges of constantly changing context and unknown event categories and struggle by designing compound text representation methods and various similarity measures. However, phased, well-designed methods are excessively fragile and unable to utilize the potential of text representations fully. Meanwhile, their complex threshold settings result in clustering-based event detection suffering the pain of ever-changing environments. We propose a text representation learning method namely Text Similarity Contrastive Learning Neural Network (Text-SimCLNN) to tackle these challenges. Text-SimCLNN uses smaller parts to learn the similarity probability of text pairs from semantic and structural perspectives, effectively bridging the gap between text representation learning and similarity measure in streaming event detection. Event discovery and merging in streams can be easily performed based on the learned representations, and we use various techniques to speed up such processes. Furthermore, we introduce an online update mechanism that uses heterogeneous graphs to generate high-quality samples to enable stable and reliable inductive learning. Extensive experiments on two real-world datasets demonstrate that our method far exceeds state-of-the-art (SOTA).

Graph Neural Networks (GNNs) have yielded fruitful results in learning multi-view graph data. However, it is challenging for existing GNNs to capture the potential correlation information (PCI) among the graph structure features of multiple views. It is also challenging to adaptively identify valuable neighbors for node feature fusion in different views. To this end, we propose a novelReinforcedTensorGraphNeuralNetwork (RTGNN) framework to more effectively perform multi-view graph representation learning through reinforcing inter- and intra-graph aggregation. Specifically, RTGNN first uses tensor decomposition to extract the graph structure features (GSFs) of each view in the common feature space. These GSFs contain the PCI of multiple views and alleviate fusion conflicts that may be caused by differences between view feature spaces in cross-view feature fusion. Since fusing the features of all neighbor nodes may harm the features of the center node, we filter the irrelevant neighbors to improve the performance of intra-graph aggregation in each view. Concretely, a reinforcement learning (RL)-guided scheme is developed to automatically calculate the optimal filtering threshold for each view, avoiding tedious manual updates and infeasible back propagation updates. Experimental results and analysis on five datasets show that RTGNN surpasses the best multi-view graph representation baselines and achieves the maximum 14.26% performance improvement in terms of F1. The code link ishttps://github.com/RingBDStack/RTGNN.

Anchor link prediction focuses on finding accounts related to the same natural person in different online platforms and can benefit several services like user modeling and cross-network recommendation systems. Recently, community structure information has attracted interest from researchers, while most of existing methods only treat community structure information as additional information and lack utilizing the interaction between communities. To address these limitations, we propose a cross-network community sensing model for anchor link prediction (CCALP). Our CCALP model regards communities as special nodes in each network and further models community level inter-network relationships by existing anchor links. We further utilize an attention mechanism to integrate cross-network community neighbors' vectors. This attention mechanism can help us balance the degree of utilization of community knowledge. Through extensive experiments on two real-world datasets, we demonstrate that CCALP outperforms the existing baseline approaches.

Recently, language models (LMs) or language representation models are widely used in natural language understanding (NLU) tasks.However, these LMs are usually trained on large unlabeled text corpora, while the finetuning process simply takes words or wordpieces as model input.Because of the differences between language model and NLU task objectives, the problem of lack of concern on some key words exists.Thus in this paper, we propose a method called reverse mapping bytepair encoding, which maps named-entity information and other word-level linguistic features back to subwords during the encoding procedure of bytepair encoding (BPE).We employ this method to the Generative Pre-trained Transformer (OpenAI GPT) (Radford et al., 2018) by adding a weighted linear layer after the embedding layer.We also propose a new model architecture named as the multi-channel separate transformer to evaluate the effectiveness of the newly introduced information by employing a training process without parameter-sharing.Experiments on Story Cloze, RTE, SciTail and SST-2 datasets demonstrate the effectiveness of our approach.Compared with the original results in GPT, our approach gains 1.58% absolute increase on Stories Cloze, 6.4% on RTE, 0.69% on SciTail and 0.8% on SST-2.

Summary Clusters based on hybrid architectures combining ARM CPUs and FPGA fabric, such as the Xilinx Zynq SoC, not only are energy‐efficient platforms with strong processing power but also have the advantages of distributed computing systems balancing the workload between cores, processors, and nodes. This paper employs a 48‐node cluster infrastructure based on the Xilinx Zynq SoC to accelerate classical cryptographic algorithms, including hash functions, AES, and RSA. In this design, we leverage the flexibility of the software to implement node‐to‐node communication through the message passing interface (MPI), and we offload the compute‐intensive tasks to the FPGA to accelerate complex calculations with the parallelizability of specific reconfigurable coprocessors. In addition, we study several parallel cryptography optimizations based on FPGA to evaluate this cluster. Finally, using a comparison with a multi‐core desktop (Intel i7‐3770) and a many‐core server (288 cores), the efficiency of the implementations of the selected data encryption and decryption algorithms is presented to illustrate the performance of our system; we also gain up to 3.6× increase in energy efficiency.

State-of-the-art Network Intrusion Detection Systems (NIDSs) use regular expressions to detect attacks or vulnerabilities. In order to keep up with the ever-increasing speed, more and more NIDSs need to be implemented by dedicated hardware. A major bottleneck is that NIDSs scan incoming packets just byte by byte, which greatly limits their throughput. In this paper, we propose a novel architecture for regular expression (RE) matching that consumes multiple characters per time. This architecture contains all the advantages of three FPGA-based algorithms to improve RE matching speed: Simple State Merge Tree (SSMT), Distribute Data in Round-Robin (DDRR), and Multi-path Speculation. Our architecture was tested on several real-life RE rulesets. It could yield a performance of 140Gbps processing rates on a single FPGA chip, while maintaining memory efficiency. This makes it a very practical solution for NIDS in 100G Ethernet standard network, which is currently the fastest approved standard of Ethernet. The experimental results also show that the throughput is about 108 times better than that of the original DFA, while the memory consumption is only about110of the original DFA.

Cooperation of software and hardware with hybrid architectures, such as Xilinx Zynq SoC combining ARM CPU and FPGA fabric, is a high-performance and low-power platform for accelerating RSA Algorithm. This paper adopts the none-subtraction Montgomery algorithm and the Chinese Remainder Theorem (CRT) to implement high-speed RSA processors, and deploys a 48-node cluster infrastructure based on Zynq SoC to achieve extremely high scalability and throughput of RSA computing. In this design, we use the ARM to implement node-to-node communication with the Message Passing Interface (MPI) while use the FPGA to handle complex calculation. Finally, the experimental results show that the overall performance is linear with the number of nodes. And the cluster achieves 6×~9× speedup against a multi-core desktop (Intel i7-3770) and comparable performance to a many-core server (288-core). In addition, we gain up to 2.5× energy efficiency compared to these two traditional platforms.

Summary Regular expression matching as a core component of deep packet inspection is widely used in various kinds of modern network intrusion detection system, traffic classification system, network monitoring system, and so on. In these systems, regular expressions are typically converted to a deterministic finite automaton (DFA), which takes O(1) to scan each input character. However, DFA generally consumes a large amount of memory. This paper proposes a novel, space‐efficient and time‐efficient DFA presentation, called reduced input character set DFA (RICS‐DFA). A character escaping and replacing scheme is first introduced to decrease the size of DFA's character set and then to reduce DFA's space requirement with a series of optimization techniques. Based on transition rewriting, a RICS‐DFA constructing algorithm with time complexity of O(n) is presented in this paper. For real rule‐sets, RICS‐DFA reduces the memory consumption by 68–92%, compared with the original DFA. Finally, this paper designs a scalable RICS‐DFA matching engine on field‐programmable gate array platform in which the reduced state transition matrix is mapped to on‐chip memories. The throughput of executing deep packet inspection for real rule‐sets can achieve 7–50.5 Gbps. Copyright © 2016 John Wiley & Sons, Ltd.

DPI technology has been widely deployed in networking intrusion detection system (NIDS) to detect attacks or viruses. State-of-the-art NIDS uses deterministic finite automata (DFA) algorithms to perform regular expression matching for its stable matching speed. However, traditional DFA algorithm's throughput is limited by the input character's width (usually one character per time). Although the multi-stride method (process multiple characters per time) can increase the throughput, it leads the DFA transition table to an exponentially increased memory consumption. In this paper, we propose a novel multi-stride regular expression matching engine called PiDFA based on Field-Programmable Gate Array (FPGA). It applies two methods to solve traditional multi-stride algorithms' memory explosion problem: DFA Transition Merging method and top-k state extraction method. Experiment results show that PiDFA achieves more than 30-fold better performance than original DFA algorithm. Whats more, PiDFA is orthogonal to existing transition table compression algorithms. Implemented with PiDFA algorithm, ClusterFA's matching speed is increased by 6-50 times while maintaining ClusterFA's low memory consumption.

The ability of ultra-low latency to process market data feed is the premise and foundation for a today's trading system to grab the instant trading profits. The market data feed containing up-to-date information on market changes is multicasted real-timely from financial exchanges to market participants, usually in the form of financial information exchange (FIX) Adapted for STreaming (FAST) protocol. FAST is a differential compression protocol which significantly reduces the bandwidth requirement to transmit market data. However, it also increases the complexity and latency of market data processing. This paper describes a customized architecture for ultra-low latency of market-data processing. Firstly, we propose a bus-based architecture of market-data decoding on Field Programmable Gate Array (FPGA). Our design is a loose-coupled and scalable architecture which is easy to adapt to different FAST templates by connecting different decoders to the main bus. Then we further exploit a dedicated pipelined design to improve the architecture. The pipelined architecture decompresses multiple messages in parallel, overcoming the challenge of data dependency between consecutive differential encoded (FAST) messages. Finally, we implement two prototypes in RTL code and evaluate them on a Xilinx Kintex-7 FPGA. Real test results show that 1) the pipelined processor gains 180% speedup compared with the non-pipelined processor; 2) it achieves an ultra-low decoding latency of 307 ns per message, which is 2 orders of magnitude faster than the software solution.

Digital right protection in P2P systems is attracting more and more attentions. In this paper, we present a new stochastic-encryption-based Digital Rights Management (DRM) scheme for P2P content delivery networks. The files are encrypted such that unpaid users cannot access the plaintext content. We exploit the random characteristics of P2P to increase the key space, which can defense collusion attacks. We add piece validation policy during a download process to prevent poisoning attacks. In our scheme, peers make a payment after downloading, and this prevents user loss due to download failures (caused by the dynamics of P2P). Our scheme does not have frequent user authentications or state maintenance. Analysis and simulation experiments show that our scheme can defend against collusion attacks and poisoning attacks with a fairly high probability.

Regular expression matching is considered important as it lies at the heart of many networking applications using deep packet inspection (DPI) techniques. For example, modern networking intrusion detection systems (NIDSs) typically accomplish regular expression matching using deterministic finite automata (DFA) algorithm. However, DFA suffers from the high memory consumption for the state blowup problem. Many algorithms have been proposed to compress the DFA memory storage space, meanwhile, they usually pay the price of low matching speed and high memory bandwidth. In this paper, we first propose an effective DFA compression algorithm by exploiting the similarity between DFA states. Then, we apply a next-state prediction strategy and present a fast DFA matching engine. Carefully designing the DFA matching circuit, we keep the prediction success rate by more than 99,5%, thus get a comparable matching speed with original DFA algorithm. On the side of memory consumption, experimental results show that with typical NIDS rule sets, our algorithm compressed the original DFA by more than 99%. Mapping this algorithm on Xilinx Virtex-7 FPGA chip, we get a throughput of more than 200Gbps.